1 2020-02-12 Jeff Law <law@redhat.com>
3 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
4 expanders, splits, etc.
5 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
6 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
7 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
8 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
9 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
10 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
12 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
14 2020-02-12 Jakub Jelinek <jakub@redhat.com>
17 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
18 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
19 TARGET_AVX512DQ from condition.
20 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
21 instead of <mask_mode512bit_condition> in condition. If
22 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
24 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
27 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
30 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
32 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
34 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
35 where strlen is more legible.
36 (rs6000_builtin_vectorized_libmass): Ditto.
37 (rs6000_print_options_internal): Ditto.
39 2020-02-11 Martin Sebor <msebor@redhat.com>
41 PR tree-optimization/93683
42 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
44 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
46 * config/rs6000/predicates.md (cint34_operand): Rename the
47 -mprefixed-addr option to be -mprefixed.
48 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
49 the -mprefixed-addr option to be -mprefixed.
50 (OTHER_FUTURE_MASKS): Likewise.
51 (POWERPC_MASKS): Likewise.
52 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
53 the -mprefixed-addr option to be -mprefixed. Change error
54 messages to refer to -mprefixed.
55 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
57 (rs6000_legitimate_offset_address_p): Likewise.
58 (rs6000_mode_dependent_address): Likewise.
59 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
60 "-mprefixed" for target attributes and pragmas.
61 (address_to_insn_form): Rename the -mprefixed-addr option to be
63 (rs6000_adjust_insn_length): Likewise.
64 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
65 -mprefixed-addr option to be -mprefixed.
66 (ASM_OUTPUT_OPCODE): Likewise.
67 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
68 -mprefixed-addr option to be -mprefixed.
69 * config/rs6000/rs6000.opt (-mprefixed): Rename the
70 -mprefixed-addr option to be prefixed. Change the option from
71 being undocumented to being documented.
72 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
73 -mprefixed option. Update the -mpcrel documentation to mention
76 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
78 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
79 including FIRST_PSEUDO_REGISTER - 1.
80 * ira-color.c (print_hard_reg_set): Ditto.
82 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
84 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
85 (USTERNOP_QUALIFIERS): New define.
86 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
87 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
88 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
89 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
90 * config/arm/arm_neon.h (vusdot_s32): New.
91 (vusdot_lane_s32): New.
92 (vusdotq_lane_s32): New.
93 (vsudot_lane_s32): New.
94 (vsudotq_lane_s32): New.
95 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
96 * config/arm/iterators.md (DOTPROD_I8MM): New.
97 (sup, opsuffix): Add <us/su>.
98 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
99 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
101 2020-02-11 Richard Biener <rguenther@suse.de>
103 PR tree-optimization/93661
104 PR tree-optimization/93662
105 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
107 * tree-sra.c (get_access_for_expr): Likewise.
109 2020-02-10 Jakub Jelinek <jakub@redhat.com>
112 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
113 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
114 Change condition from TARGET_AVX2 to TARGET_AVX.
116 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
119 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
122 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
124 Try to generate zero-based comparisons.
125 * config/cris/cris.c (cris_reduce_compare): New function.
126 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
127 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
128 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
130 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
133 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
134 in Thumb state and also as a destination in Arm state. Add T16
137 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
139 * md.texi (Define Subst): Match closing paren in example.
141 2020-02-10 Jakub Jelinek <jakub@redhat.com>
145 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
146 arguments of strncmp.
148 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
151 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
152 but different source value.
153 (adjust_callers_for_value_intersection): New function.
154 (gather_edges_for_value): Adjust order of callers to let a
155 non-self-recursive caller be the first element.
156 (self_recursive_pass_through_p): Add a new parameter "simple", and
157 check generalized self-recursive pass-through jump function.
158 (self_recursive_agg_pass_through_p): Likewise.
159 (find_more_scalar_values_for_callers_subset): Compute value from
160 pass-through jump function for self-recursive.
161 (intersect_with_plats): Cleanup previous implementation code for value
162 itersection with self-recursive call edge.
163 (intersect_with_agg_replacements): Likewise.
164 (intersect_aggregates_with_edge): Deduce value from pass-through jump
165 function for self-recursive call edge. Cleanup previous implementation
166 code for value intersection with self-recursive call edge.
167 (decide_whether_version_node): Remove dead callers and adjust order
168 to let a non-self-recursive caller be the first element.
170 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
172 * recog.c: Move pass_split_before_sched2 code in front of
173 pass_split_before_regstack.
174 (pass_data_split_before_sched2): Rename pass to split3 from split4.
175 (pass_data_split_before_regstack): Rename pass to split4 from split3.
176 (rest_of_handle_split_before_sched2): Remove.
177 (pass_split_before_sched2::execute): Unconditionally call
179 (enable_split_before_sched2): New function.
180 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
181 (pass_split_before_regstack::gate): Ditto.
182 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
183 Update name check for renamed split4 pass.
184 * config/sh/sh.c (register_sh_passes): Update pass insertion
185 point for renamed split4 pass.
187 2020-02-09 Jakub Jelinek <jakub@redhat.com>
189 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
190 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
191 copying them around between host and target.
193 2020-02-08 Andrew Pinski <apinski@marvell.com>
196 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
197 STRICT_ALIGNMENT also.
199 2020-02-08 Jim Wilson <jimw@sifive.com>
202 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
204 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
205 Jakub Jelinek <jakub@redhat.com>
208 * config/i386/i386.h (CALL_USED_REGISTERS): Make
209 xmm16-xmm31 call-used even in 64-bit ms-abi.
211 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
213 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
214 (simd_ummla, simd_usmmla): Likewise.
215 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
216 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
219 2020-02-07 Richard Biener <rguenther@suse.de>
222 * tree-inline.c (fold_marked_statements): Do a PRE walk,
223 skipping unreachable regions.
224 (optimize_inline_calls): Skip folding stmts when we didn't
227 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
230 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
231 Don't return aggregates with only SFmode and DFmode in SSE
233 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
235 2020-02-07 Jakub Jelinek <jakub@redhat.com>
238 * config/rs6000/rs6000-logue.c
239 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
240 if it fails, move rs into end_addr and retry. Add
241 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
242 the insn pattern doesn't describe well what exactly happens to
246 * config/i386/predicates.md (avx_identity_operand): Remove.
247 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
248 (avx_<castmode><avxsizesuffix>_<castmode>,
249 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
250 a VEC_CONCAT of the operand and UNSPEC_CAST.
251 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
252 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
256 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
257 recog_data.insn if distance_non_agu_define changed it.
259 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
262 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
263 we only had X-FORM (reg+reg) addressing for vectors. Also before
264 ISA 3.0, we only had X-FORM addressing for scalars in the
265 traditional Altivec registers.
267 2020-02-06 <zhongyunde@huawei.com>
268 Vladimir Makarov <vmakarov@redhat.com>
270 PR rtl-optimization/93561
271 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
274 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
276 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
279 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
281 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
282 where the low and the high 32 bits are equal to each other specially,
283 with an rldimi instruction.
285 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
287 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
289 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
291 * config/arm/arm-tables.opt: Regenerate.
293 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
296 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
297 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
298 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
300 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
302 PR rtl-optimization/87763
303 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
305 2020-02-06 Delia Burduv <delia.burduv@arm.com>
307 * config/aarch64/aarch64-simd-builtins.def
308 (bfmlaq): New built-in function.
309 (bfmlalb): New built-in function.
310 (bfmlalt): New built-in function.
311 (bfmlalb_lane): New built-in function.
312 (bfmlalt_lane): New built-in function.
313 * config/aarch64/aarch64-simd.md
314 (aarch64_bfmmlaqv4sf): New pattern.
315 (aarch64_bfmlal<bt>v4sf): New pattern.
316 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
317 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
318 (vbfmlalbq_f32): New intrinsic.
319 (vbfmlaltq_f32): New intrinsic.
320 (vbfmlalbq_lane_f32): New intrinsic.
321 (vbfmlaltq_lane_f32): New intrinsic.
322 (vbfmlalbq_laneq_f32): New intrinsic.
323 (vbfmlaltq_laneq_f32): New intrinsic.
324 * config/aarch64/iterators.md (BF_MLA): New int iterator.
325 (bt): New int attribute.
327 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
329 * config/i386/i386.md (*pushtf): Emit "#" instead of
330 calling gcc_unreachable in insn output.
333 (*pushsf_rex64): Ditto for alternatives other than 1.
334 (*pushsf): Ditto for alternatives other than 1.
336 2020-02-06 Martin Liska <mliska@suse.cz>
338 PR gcov-profile/91971
339 PR gcov-profile/93466
340 * coverage.c (coverage_init): Revert mangling of
341 path into filename. It can lead to huge filename length.
342 Creation of subfolders seem more natural.
344 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
347 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
348 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
349 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
351 2020-02-06 Jakub Jelinek <jakub@redhat.com>
354 * config/i386/predicates.md (avx_identity_operand): New predicate.
355 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
356 define_insn_and_split.
359 * omp-low.c (use_pointer_for_field): For nested constructs, also
360 look for map clauses on target construct.
361 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
362 taskreg_nesting_level.
365 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
366 shared clause, call omp_notice_variable on outer context if any.
368 2020-02-05 Jason Merrill <jason@redhat.com>
371 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
372 non-zero address even if weak and not yet defined.
374 2020-02-05 Martin Sebor <msebor@redhat.com>
376 PR tree-optimization/92765
377 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
378 * tree-ssa-strlen.c (compute_string_length): Remove.
379 (determine_min_objsize): Remove.
380 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
381 Avoid using type size as the upper bound on string length.
382 (handle_builtin_string_cmp): Add an argument. Adjust.
383 (strlen_check_and_optimize_call): Pass additional argument to
384 handle_builtin_string_cmp.
386 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
388 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
389 (*pushdi2_rex64 peephole2): Unconditionally split after
391 (*ashl<mode>3_doubleword): Ditto.
392 (*<shift_insn><mode>3_doubleword): Ditto.
394 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
397 * config/rs6000/rs6000.c (get_vector_offset): Fix
399 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
401 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
403 2020-02-05 David Malcolm <dmalcolm@redhat.com>
406 (Special Functions for Debugging the Analyzer): Update description
407 of __analyzer_dump_exploded_nodes.
409 2020-02-05 Jakub Jelinek <jakub@redhat.com>
412 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
413 include sets and not clobbers in the vzeroupper pattern.
414 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
415 the parallel has 17 (64-bit) or 9 (32-bit) elts.
416 (*avx_vzeroupper_1): New define_insn_and_split.
419 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
420 don't run when !optimize.
421 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
424 2020-02-05 Richard Biener <rguenther@suse.de>
427 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
428 checks before matching calls.
430 2020-02-05 Jakub Jelinek <jakub@redhat.com>
432 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
433 function comment typo.
436 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
437 simd_clone_create failed when i == 0, adjust clone->nargs by
440 2020-02-05 Martin Liska <mliska@suse.cz>
443 * doc/invoke.texi: Document that one should
444 not combine ASLR and -fpch.
446 2020-02-04 Richard Biener <rguenther@suse.de>
448 PR tree-optimization/93538
449 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
451 2020-02-04 Richard Biener <rguenther@suse.de>
453 PR tree-optimization/91123
454 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
455 (vn_walk_cb_data::last_vuse): New member.
456 (vn_walk_cb_data::saved_operands): Likewsie.
457 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
458 (vn_walk_cb_data::push_partial_def): Use finish.
459 (vn_reference_lookup_2): Update last_vuse and use finish if
460 we've saved operands.
461 (vn_reference_lookup_3): Use finish and update calls to
462 push_partial_defs everywhere. When translating through
463 memcpy or aggregate copies save off operands and alias-set.
464 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
465 operation for redundant store removal.
467 2020-02-04 Richard Biener <rguenther@suse.de>
469 PR tree-optimization/92819
470 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
471 generating more stmts than before.
473 2020-02-04 Martin Liska <mliska@suse.cz>
475 * config/arm/arm.c (arm_gen_far_branch): Move the function
476 outside of selftests.
478 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
480 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
481 function to adjust PC-relative vector addresses.
482 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
483 handle vectors with PC-relative addresses.
485 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
487 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
489 (hard_reg_and_mode_to_addr_mask): Delete.
490 (rs6000_adjust_vec_address): If the original vector address
491 was REG+REG or REG+OFFSET and the element is not zero, do the add
492 of the elements in the original address before adding the offset
493 for the vector element. Use address_to_insn_form to validate the
494 address using the register being loaded, rather than guessing
495 whether the address is a DS-FORM or DQ-FORM address.
497 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
499 * config/rs6000/rs6000.c (get_vector_offset): New helper function
500 to calculate the offset in memory from the start of a vector of a
501 particular element. Add code to keep the element number in
502 bounds if the element number is variable.
503 (rs6000_adjust_vec_address): Move calculation of offset of the
504 vector element to get_vector_offset.
505 (rs6000_split_vec_extract_var): Do not do the initial AND of
506 element here, move the code to get_vector_offset.
508 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
510 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
513 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
515 * config/rs6000/constraints.md: Improve documentation.
517 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
520 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
521 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
523 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
525 * config.gcc: Remove "carrizo" support.
526 * config/gcn/gcn-opts.h (processor_type): Likewise.
527 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
528 * config/gcn/gcn.opt (gpu_type): Likewise.
529 * config/gcn/t-omp-device: Likewise.
531 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
534 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
535 * config/arm/arm.c (arm_gen_far_branch): New function
537 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
539 2020-02-03 Julian Brown <julian@codesourcery.com>
540 Tobias Burnus <tobias@codesourcery.com>
542 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
544 2020-02-03 Jakub Jelinek <jakub@redhat.com>
547 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
548 valid RTL to sum up the lowest and second lowest bytes of the popcnt
551 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
553 PR rtl-optimization/91333
554 * ira-color.c (struct allocno_color_data): Add member
556 (init_allocno_threads): Set the member up.
557 (bucket_allocno_compare_func): Add compare hard reg
560 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
562 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
564 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
565 * config.in: Regenerated.
566 * configure: Regenerated.
567 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
568 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
569 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
571 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
573 * configure: Regenerate.
575 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
577 PR rtl-optimization/91333
578 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
579 reg preferences comparison up.
581 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
583 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
584 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
585 aarch64-sve-builtins-base.h.
586 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
587 aarch64-sve-builtins-base.cc.
588 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
589 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
591 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
592 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
593 (svcvtnt): New functions.
594 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
595 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
596 (svcvtnt): New functions.
597 (svcvt): Add a form that converts f32 to bf16.
598 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
599 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
601 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
602 Treat B as bfloat16_t.
603 (ternary_bfloat_lane_base): New class.
604 (ternary_bfloat_def): Likewise.
605 (ternary_bfloat): New shape.
606 (ternary_bfloat_lane_def): New class.
607 (ternary_bfloat_lane): New shape.
608 (ternary_bfloat_lanex2_def): New class.
609 (ternary_bfloat_lanex2): New shape.
610 (ternary_bfloat_opt_n_def): New class.
611 (ternary_bfloat_opt_n): New shape.
612 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
613 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
614 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
615 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
616 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
617 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
618 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
619 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
620 the pattern off the narrow mode instead of the wider one.
621 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
622 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
623 (sve_fp_op): Handle them.
624 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
625 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
627 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
629 * config/aarch64/arm_sve.h: Include arm_bf16.h.
630 * config/aarch64/aarch64-modes.def (BF): Move definition before
631 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
632 (SVE_MODES): Handle BF modes.
633 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
635 (aarch64_full_sve_mode): Likewise.
636 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
638 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
639 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
640 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
641 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
643 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
645 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
646 (TYPES_all_data): Add bf16.
647 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
648 (register_tuple_type): Increase buffer size.
649 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
650 (bf16): New type suffix.
651 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
652 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
653 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
654 Change type from all_data to all_arith.
655 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
658 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
659 Matthew Malcomson <matthew.malcomson@arm.com>
660 Richard Sandiford <richard.sandiford@arm.com>
662 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
663 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
664 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
665 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
666 __ARM_FEATURE_MATMUL_FP64.
667 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
668 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
669 be disabled at the same time.
670 (f32mm): New extension.
671 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
672 (AARCH64_FL_F64MM): Bump to the next bit up.
673 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
674 (TARGET_SVE_F64MM): New macros.
675 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
676 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
677 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
678 (UNSPEC_ZIP2Q): New unspeccs.
679 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
680 (optab, sur, perm_insn): Handle the new unspecs.
681 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
682 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
683 TARGET_SVE_F64MM instead of separate tests.
684 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
685 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
686 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
687 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
688 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
689 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
690 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
691 (TYPES_s_signed): New macro.
692 (TYPES_s_integer): Use it.
693 (TYPES_d_float): New macro.
694 (TYPES_d_data): Use it.
695 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
696 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
697 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
698 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
700 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
702 (ternary_resize2_lane_base): Likewise.
703 (ternary_resize2_base): New class.
704 (ternary_qq_lane_base): Likewise.
705 (ternary_intq_uintq_lane_def): Likewise.
706 (ternary_intq_uintq_lane): New shape.
707 (ternary_intq_uintq_opt_n_def): New class
708 (ternary_intq_uintq_opt_n): New shape.
709 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
710 (ternary_uintq_intq_def): New class.
711 (ternary_uintq_intq): New shape.
712 (ternary_uintq_intq_lane_def): New class.
713 (ternary_uintq_intq_lane): New shape.
714 (ternary_uintq_intq_opt_n_def): New class.
715 (ternary_uintq_intq_opt_n): New shape.
716 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
717 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
718 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
719 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
721 (svdotprod_lane_impl): ...this new class.
722 (svmmla_impl, svusdot_impl): New classes.
723 (svdot_lane): Update to use svdotprod_lane_impl.
724 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
725 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
727 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
728 function, with no types defined.
729 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
730 AARCH64_FL_I8MM functions.
731 (svmmla): New AARCH64_FL_F32MM function.
732 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
733 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
734 AARCH64_FL_F64MM function.
735 (REQUIRED_EXTENSIONS):
737 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
739 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
742 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
744 * config/i386/i386.md (*movoi_internal_avx): Do not check for
745 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
746 (*movti_internal): Do not check for
747 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
748 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
749 just after check for TARGET_AVX.
750 (*movdf_internal): Ditto.
751 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
752 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
753 * config/i386/sse.md (mov<mode>_internal): Only check
754 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
755 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
756 (<sse>_andnot<mode>3<mask_name>): Move check for
757 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
758 (<code><mode>3<mask_name>): Ditto.
759 (*andnot<mode>3): Ditto.
761 (*<code><mode>3): Ditto.
763 (*andnot<VI:mode>3): Remove
764 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
765 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
766 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
767 (sse4_1_blendv<ssemodesuffix>): Ditto.
768 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
769 Explain that tune applies to 128bit instructions only.
771 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
773 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
774 to definition of hsa_kernel_description. Parse assembly to find SGPR
775 and VGPR count of kernel and store in hsa_kernel_description.
777 2020-01-31 Tamar Christina <tamar.christina@arm.com>
779 PR rtl-optimization/91838
780 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
781 to truncate if allowed or reject combination.
783 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
785 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
786 (find_inv_vars_cb): Likewise.
788 2020-01-31 David Malcolm <dmalcolm@redhat.com>
790 * calls.c (special_function_p): Split out the check for DECL_NAME
791 being non-NULL and fndecl being extern at file scope into a
792 new maybe_special_function_p and call it. Drop check for fndecl
793 being non-NULL that was after a usage of DECL_NAME (fndecl).
794 * tree.h (maybe_special_function_p): New inline function.
796 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
798 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
799 (mask_gather_load<mode>): ... here, and zero-initialize the
801 (maskload<mode>di): Zero-initialize the destination.
804 2020-01-30 David Malcolm <dmalcolm@redhat.com>
807 * doc/analyzer.texi (Limitations): Note that constraints on
808 floating-point values are currently ignored.
810 2020-01-30 Jakub Jelinek <jakub@redhat.com>
813 * symtab.c (symtab_node::noninterposable_alias): If localalias
814 already exists, but is not usable, append numbers after it until
815 a unique name is found. Formatting fix.
818 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
821 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
823 * config/gcn/gcn.c (print_operand): Handle LTGT.
824 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
826 2020-01-30 Richard Biener <rguenther@suse.de>
828 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
829 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
831 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
833 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
834 without a DECL in .data.rel.ro.local.
836 2020-01-30 Jakub Jelinek <jakub@redhat.com>
839 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
844 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
845 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
846 any_extend code iterator instead of always zero_extend.
847 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
848 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
849 Use any_extend code iterator instead of always zero_extend.
850 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
851 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
852 Use any_extend code iterator instead of always zero_extend.
853 (*sse2_pmovmskb_ext): New define_insn.
854 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
857 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
858 (*popcountsi2_zext_falsedep): New define_insn.
860 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
862 * config.in: Regenerated.
863 * configure: Regenerated.
865 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
868 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
869 LLVM's assembler changed the default in version 9.
871 2020-01-24 Jeff Law <law@redhat.com>
873 PR tree-optimization/89689
874 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
876 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
880 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
882 PR rtl-optimization/87763
883 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
884 simplification to handle subregs as well as bare regs.
885 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
887 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
890 * ira.c (ira): Revert use of simplified LRA algorithm.
892 2020-01-29 Martin Jambor <mjambor@suse.cz>
894 PR tree-optimization/92706
895 * tree-sra.c (struct access): Fields first_link, last_link,
896 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
897 next_rhs_queued and grp_rhs_queued respectively, new fields
898 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
899 (struct assign_link): Field next renamed to next_rhs, new field
900 next_lhs. Updated comment.
901 (work_queue_head): Renamed to rhs_work_queue_head.
902 (lhs_work_queue_head): New variable.
903 (add_link_to_lhs): New function.
904 (relink_to_new_repr): Also relink LHS lists.
905 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
906 (add_access_to_lhs_work_queue): New function.
907 (pop_access_from_work_queue): Renamed to
908 pop_access_from_rhs_work_queue.
909 (pop_access_from_lhs_work_queue): New function.
910 (build_accesses_from_assign): Also add links to LHS lists and to LHS
912 (child_would_conflict_in_lacc): Renamed to
913 child_would_conflict_in_acc. Adjusted parameter names.
914 (create_artificial_child_access): New parameter set_grp_read, use it.
915 (subtree_mark_written_and_enqueue): Renamed to
916 subtree_mark_written_and_rhs_enqueue.
917 (propagate_subaccesses_across_link): Renamed to
918 propagate_subaccesses_from_rhs.
919 (propagate_subaccesses_from_lhs): New function.
920 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
923 2020-01-29 Martin Jambor <mjambor@suse.cz>
925 PR tree-optimization/92706
926 * tree-sra.c (struct access): Adjust comment of
927 grp_total_scalarization.
928 (find_access_in_subtree): Look for single children spanning an entire
930 (scalarizable_type_p): Allow register accesses, adjust callers.
931 (completely_scalarize): Remove function.
932 (scalarize_elem): Likewise.
933 (create_total_scalarization_access): Likewise.
934 (sort_and_splice_var_accesses): Do not track total scalarization
936 (analyze_access_subtree): New parameter totally, adjust to new meaning
937 of grp_total_scalarization.
938 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
939 (can_totally_scalarize_forest_p): New function.
940 (create_total_scalarization_access): Likewise.
941 (create_total_access_and_reshape): Likewise.
942 (total_should_skip_creating_access): Likewise.
943 (totally_scalarize_subtree): Likewise.
944 (analyze_all_variable_accesses): Perform total scalarization after
945 subaccess propagation using the new functions above.
946 (initialize_constant_pool_replacements): Output initializers by
947 traversing the access tree.
949 2020-01-29 Martin Jambor <mjambor@suse.cz>
951 * tree-sra.c (verify_sra_access_forest): New function.
952 (verify_all_sra_access_forests): Likewise.
953 (create_artificial_child_access): Set parent.
954 (analyze_all_variable_accesses): Call the verifier.
956 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
958 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
959 if called on indirect edge.
960 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
961 speculative call if needed.
963 2020-01-29 Richard Biener <rguenther@suse.de>
965 PR tree-optimization/93428
966 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
967 permutation when the load node is created.
968 (vect_analyze_slp_instance): Re-use it here.
970 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
972 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
974 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
976 PR rtl-optimization/93272
977 * ira-lives.c (process_out_of_region_eh_regs): New function.
978 (process_bb_node_lives): Call it.
980 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
982 * coverage.c (read_counts_file): Make error message lowercase.
984 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
986 * profile-count.c (profile_quality_display_names): Fix ordering.
988 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
991 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
992 hash only when edge is first within the sequence.
993 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
994 (symbol_table::create_edge): Do not set target_prob.
995 (cgraph_edge::remove_caller): Watch for speculative calls when updating
997 (cgraph_edge::make_speculative): Drop target_prob parameter.
998 (cgraph_edge::speculative_call_info): Remove.
999 (cgraph_edge::first_speculative_call_target): New member function.
1000 (update_call_stmt_hash_for_removing_direct_edge): New function.
1001 (cgraph_edge::resolve_speculation): Rewrite to new API.
1002 (cgraph_edge::speculative_call_for_target): New member function.
1003 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1004 multiple speculation targets.
1005 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1007 (verify_speculative_call): Verify that targets form an interval.
1008 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1009 (cgraph_edge::first_speculative_call_target): New member function.
1010 (cgraph_edge::next_speculative_call_target): New member function.
1011 (cgraph_edge::speculative_call_target_ref): New member function.
1012 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1013 (cgraph_edge): Remove target_prob.
1014 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1015 Fix handling of speculative calls.
1016 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1017 * ipa-fnsummary.c (analyze_function_body): Likewise.
1018 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1019 * ipa-profile.c (dump_histogram): Fix formating.
1020 (ipa_profile_generate_summary): Watch for overflows.
1021 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1022 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1023 (update_indirect_edges_after_inlining): Update to new API.
1024 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1026 * profile-count.h: (profile_probability::adjusted): New.
1027 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1028 updating of profile.
1029 * value-prof.c (gimple_ic_transform): Rename to ...
1030 (dump_ic_profile): ... this one; update dumping.
1031 (stream_in_histogram_value): Fix formating.
1032 (gimple_value_profile_transformations): Update.
1034 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1037 * config/i386/i386.md (*movoi_internal_avx): Remove
1038 TARGET_SSE_TYPELESS_STORES check.
1039 (*movti_internal): Prefer TARGET_AVX over
1040 TARGET_SSE_TYPELESS_STORES.
1041 (*movtf_internal): Likewise.
1042 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1043 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1044 from TARGET_SSE_TYPELESS_STORES.
1046 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1048 * diagnostic-core.h (warning_at): Rename overload to...
1049 (warning_meta): ...this.
1050 (emit_diagnostic_valist): Delete decl of overload taking
1051 diagnostic_metadata.
1052 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1053 (warning_at): Rename overload taking diagnostic_metadata to...
1054 (warning_meta): ...this.
1056 2020-01-28 Richard Biener <rguenther@suse.de>
1058 PR tree-optimization/93439
1059 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1060 * tree-cfg.c (move_sese_region_to_fn): ... here.
1061 (verify_types_in_gimple_reference): Verify used cliques are
1064 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1067 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1068 argument of a pointer to struct gcc_options and pass it to
1069 parse_mtune_ctrl_str.
1070 (ix86_function_specific_restore): Pass opts to
1071 set_ix86_tune_features.
1072 (ix86_option_override_internal): Likewise.
1073 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1074 gcc_options and use it for x_ix86_tune_ctrl_string.
1076 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1078 PR rtl-optimization/87763
1079 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1080 simplification to handle subregs as well as bare regs.
1081 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1083 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1085 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1086 for reduction chains that (now) include a call.
1088 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1090 PR tree-optimization/92822
1091 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1092 out the don't-care elements of a vector whose significant elements
1093 are duplicates, make the don't-care elements duplicates too.
1095 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1097 PR tree-optimization/93434
1098 * tree-predcom.c (split_data_refs_to_components): Record which
1099 components have had aliasing loads removed. Prevent store-store
1100 commoning for all such components.
1102 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1105 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1106 -1 or is_vshift is true, use new_vector with number of elts npatterns
1107 rather than new_unary_operation.
1109 PR tree-optimization/93454
1110 * gimple-fold.c (fold_array_ctor_reference): Perform
1111 elt_size.to_uhwi () just once, instead of calling it in every
1112 iteration. Punt if that value is above size of the temporary
1113 buffer. Decrease third native_encode_expr argument when
1114 bufoff + elt_sz is above size of buf.
1116 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1118 * config/mips/mips.c (mips_declare_object_name)
1119 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1121 2020-01-27 Martin Liska <mliska@suse.cz>
1123 PR gcov-profile/93403
1124 * tree-profile.c (gimple_init_gcov_profiler): Generate
1125 both __gcov_indirect_call_profiler_v4 and
1126 __gcov_indirect_call_profiler_v4_atomic.
1128 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1131 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1133 (@aarch64_split_simd_mov<mode>): Use it.
1134 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1135 Leave the vec_extract patterns to handle 2-element vectors.
1136 (aarch64_simd_mov_from_<mode>high): Likewise.
1137 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1138 (vec_extractv2dfv1df): Likewise.
1140 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1142 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1143 jump conditions for *compare_condjump<GPI:mode>.
1145 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1148 * digraph.cc (test_edge::test_edge): Specify template for base
1151 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1153 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1155 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1157 * config/arc/arc-protos.h (gen_mlo): Remove.
1158 (gen_mhi): Likewise.
1159 * config/arc/arc.c (AUX_MULHI): Define.
1160 (arc_must_save_reister): Special handling for r58/59.
1161 (arc_compute_frame_size): Consider mlo/mhi registers.
1162 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1164 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1165 mlo/mhi name selection.
1166 (arc_restore_callee_saves): Don't early restore blink when ISR.
1167 (arc_expand_prologue): Add mlo/mhi saving.
1168 (arc_expand_epilogue): Add mlo/mhi restoring.
1171 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1172 numbering when MUL64 option is used.
1173 (DWARF2_FRAME_REG_OUT): Define.
1174 * config/arc/arc.md (arc600_stall): New pattern.
1175 (VUNSPEC_ARC_ARC600_STALL): Define.
1176 (mulsi64): Use correct mlo/mhi registers.
1177 (mulsi_600): Clean it up.
1178 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1180 (mhi_operand): Likewise.
1182 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1183 Petro Karashchenko <petro.karashchenko@ring.com>
1185 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1186 attributes if needed.
1187 (prepare_move_operands): Generate special unspec instruction for
1189 (arc_isuncached_mem_p): Propagate uncached attribute to each
1191 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1192 (VUNSPEC_ARC_STDI): Likewise.
1193 (ALLI): New mode iterator.
1194 (mALLI): New mode attribute.
1195 (lddi): New instruction pattern.
1197 (stdidi_split): Split instruction for architectures which are not
1198 supporting ll64 option.
1199 (lddidi_split): Likewise.
1201 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1203 PR rtl-optimization/92989
1204 * lra-lives.c (process_bb_lives): Update the live-in set before
1205 processing additional clobbers.
1207 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1209 PR rtl-optimization/93170
1210 * cselib.c (cselib_invalidate_regno_val): New function, split out
1212 (cselib_invalidate_regno): ...here.
1213 (cselib_invalidated_by_call_p): New function.
1214 (cselib_process_insn): Iterate over all the hard-register entries in
1215 REG_VALUES and invalidate any that cross call-clobbered registers.
1217 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1219 * dojump.c (split_comparison): Use HONOR_NANS rather than
1220 HONOR_SNANS when splitting LTGT.
1222 2020-01-27 Martin Liska <mliska@suse.cz>
1225 * opts.c (print_filtered_help): Exclude language-specific
1226 options from --help=common unless enabled in all FEs.
1228 2020-01-27 Martin Liska <mliska@suse.cz>
1230 * opts.c (print_help): Exclude params from
1231 all except --help=param.
1233 2020-01-27 Martin Liska <mliska@suse.cz>
1236 * config/i386/i386-features.c (make_resolver_func):
1237 Align the code with ppc64 target implementation.
1238 Do not generate a unique name for resolver function.
1240 2020-01-27 Richard Biener <rguenther@suse.de>
1242 PR tree-optimization/93397
1243 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1244 converted reduction chain SLP graph adjustment.
1246 2020-01-26 Marek Polacek <polacek@redhat.com>
1249 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1252 2020-01-26 Jason Merrill <jason@redhat.com>
1255 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1258 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1260 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1261 (rx_setmem): Likewise.
1263 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1266 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1267 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1268 drop <di> from constraint of last operand.
1271 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1272 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1273 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1275 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1278 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1281 2020-01-24 Jeff Law <law@redhat.com>
1283 PR tree-optimization/92788
1284 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1287 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1290 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1291 *avx_vperm_broadcast_<mode>,
1292 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1293 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1294 Move before avx2_perm<mode>/avx512f_perm<mode>.
1297 * simplify-rtx.c (simplify_const_unary_operation,
1298 simplify_const_binary_operation): Punt for mode precision above
1299 MAX_BITSIZE_MODE_ANY_INT.
1301 2020-01-24 Andrew Pinski <apinski@marvell.com>
1303 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1306 2020-01-24 Jeff Law <law@redhat.com>
1309 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1310 for REGs. Call output_operand_lossage to get more reasonable
1313 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1315 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1316 gcn_fp_compare_operator.
1317 (vec_cmpu<mode>di): Use gcn_compare_operator.
1318 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1319 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1320 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1321 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1322 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1323 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1324 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1325 gcn_fp_compare_operator.
1326 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1327 gcn_fp_compare_operator.
1328 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1329 gcn_fp_compare_operator.
1330 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1331 gcn_fp_compare_operator.
1333 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1335 * doc/install.texi (Cross-Compiler-Specific Options): Document
1336 `--with-toolexeclibdir' option.
1338 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1340 * target.def (flags_regnum): Also mention effect on delay slot filling.
1341 * doc/tm.texi: Regenerate.
1343 2020-01-23 Jeff Law <law@redhat.com>
1345 PR translation/90162
1346 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1348 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1351 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1354 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1356 PR rtl-optimization/93402
1357 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1360 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1362 * config.in: Regenerated.
1363 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1364 for TARGET_LIBC_GNUSTACK.
1365 * configure: Regenerated.
1366 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1367 found to be 2.31 or greater.
1369 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1371 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1373 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1374 (mips_asm_file_end): New function. Delegate to
1375 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1376 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1378 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1381 * config/i386/i386-modes.def (POImode): New mode.
1382 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1383 * config/i386/i386.md (DPWI): New mode attribute.
1384 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1386 (QPWI): ... this. Use POI instead of OI for TImode.
1387 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1388 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1391 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1394 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1396 (speculation_tracker_rev): New pattern.
1397 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1398 Use speculation_tracker_rev to track the inverse condition.
1400 2020-01-23 Richard Biener <rguenther@suse.de>
1402 PR tree-optimization/93381
1403 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1404 alias-set of the def as argument and record the first one.
1405 (vn_walk_cb_data::first_set): New member.
1406 (vn_reference_lookup_3): Pass the alias-set of the current def
1407 to push_partial_def. Fix alias-set used in the aggregate copy
1409 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1410 * real.c (clear_significand_below): Fix out-of-bound access.
1412 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1415 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1416 New define_insn patterns.
1418 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1420 * doc/sourcebuild.texi (check-function-bodies): Add an
1421 optional target/xfail selector.
1423 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1425 PR rtl-optimization/93124
1426 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1427 bare USE and CLOBBER insns.
1429 2020-01-22 Andrew Pinski <apinski@marvell.com>
1431 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1433 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1436 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1437 diagnostic_manager into "ana" namespace.
1438 * selftest-run-tests.c (selftest::run_tests): Update for move of
1439 selftest::run_analyzer_selftests to
1440 ana::selftest::run_analyzer_selftests.
1442 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1444 * cfgexpand.c (union_stack_vars): Update the size.
1446 2020-01-22 Richard Biener <rguenther@suse.de>
1448 PR tree-optimization/93381
1449 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1450 throughout, handle all conversions the same.
1452 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1455 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1456 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1457 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1458 Call force_reg on high_in2 unconditionally.
1460 2020-01-22 Martin Liska <mliska@suse.cz>
1462 PR tree-optimization/92924
1463 * profile.c (compute_value_histograms): Divide
1466 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1469 * output.h (assemble_name_resolve): Declare.
1470 * varasm.c (assemble_name_resolve): New function.
1471 (assemble_name): Use it.
1472 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1474 2020-01-22 Joseph Myers <joseph@codesourcery.com>
1476 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1477 update_web_docs_git instead of update_web_docs_svn.
1479 2020-01-21 Andrew Pinski <apinski@marvell.com>
1482 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1483 as PTR mode. Have operand 1 as being modeless, it can be P mode.
1484 (*tlsgd_small_<mode>): Likewise.
1485 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1486 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1487 register. Convert that register back to dest using convert_mode.
1489 2020-01-21 Jim Wilson <jimw@sifive.com>
1491 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1494 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1495 Uros Bizjak <ubizjak@gmail.com>
1498 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1500 (legitimize_tls_address): Do GNU2 TLS address computation in
1501 ptr_mode and zero-extend result to Pmode.
1502 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1503 :P with :PTR and Pmode with ptr_mode.
1504 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1505 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1506 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1508 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1511 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1512 the last two operands are CONST_INT_P before using them as such.
1514 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1516 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1517 to get the integer element types.
1519 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1521 * config/aarch64/aarch64-sve-builtins.h
1522 (function_expander::convert_to_pmode): Declare.
1523 * config/aarch64/aarch64-sve-builtins.cc
1524 (function_expander::convert_to_pmode): New function.
1525 (function_expander::get_contiguous_base): Use it.
1526 (function_expander::prepare_gather_address_operands): Likewise.
1527 * config/aarch64/aarch64-sve-builtins-sve2.cc
1528 (svwhilerw_svwhilewr_impl::expand): Likewise.
1530 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
1533 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
1534 cfun->machine->label_is_assembled.
1535 (aarch64_print_patchable_function_entry): New.
1536 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
1537 * config/aarch64/aarch64.h (struct machine_function): New field,
1540 2020-01-21 David Malcolm <dmalcolm@redhat.com>
1543 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1546 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1549 * cgraph.c (cgraph_edge::resolve_speculation,
1550 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
1551 call_stmt_site_hash.
1553 2020-01-21 Martin Liska <mliska@suse.cz>
1555 * config/rs6000/rs6000.c (common_mode_defined): Remove
1558 2020-01-21 Richard Biener <rguenther@suse.de>
1560 PR tree-optimization/92328
1561 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
1562 type when value-numbering same-sized store by inserting a
1564 (eliminate_dom_walker::eliminate_stmt): When eliminating
1565 a redundant store handle bit-reinterpretation of the same value.
1567 2020-01-21 Andrew Pinski <apinski@marvel.com>
1570 * tree-into-ssa.c (prepare_block_for_update_1): Split out
1572 (prepare_block_for_update): This. Use a worklist instead of
1575 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1577 * gcc/config/arm/arm.c (clear_operation_p):
1578 Initialise last_regno, skip first iteration
1579 based on the first_set value and use ints instead
1580 of the unnecessary HOST_WIDE_INTs.
1582 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1585 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
1586 compare_mode other than SFmode or DFmode.
1588 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1591 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
1592 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
1593 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
1595 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1597 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
1599 2020-01-20 Andrew Pinski <apinski@marvell.com>
1602 * targhooks.c (default_print_patchable_function_entry): Use
1603 output_asm_insn to emit the nop instruction.
1605 2020-01-20 Fangrui Song <maskray@google.com>
1608 * targhooks.c (default_print_patchable_function_entry): Align to
1611 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
1614 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
1615 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
1616 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
1617 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
1618 (*tls_dynamic_gnu2_lea_64): Renamed to ...
1619 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
1620 Remove the {q} suffix from lea.
1621 (*tls_dynamic_gnu2_call_64): Renamed to ...
1622 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
1623 (*tls_dynamic_gnu2_combine_64): Renamed to ...
1624 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
1625 Pass Pmode to gen_tls_dynamic_gnu2_64.
1627 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1629 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
1631 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
1633 * config/aarch64/aarch64-sve-builtins-base.cc
1634 (svld1ro_impl::memory_vector_mode): Remove parameter name.
1636 2020-01-20 Richard Biener <rguenther@suse.de>
1639 * dwarf2out.c (prune_unused_types): Unconditionally mark
1640 called function DIEs.
1642 2020-01-20 Martin Liska <mliska@suse.cz>
1644 PR tree-optimization/93199
1645 * tree-eh.c (struct leh_state): Add
1646 new field outer_non_cleanup.
1647 (cleanup_is_dead_in): Pass leh_state instead
1648 of eh_region. Add a checking that state->outer_non_cleanup
1649 points to outer non-clean up region.
1650 (lower_try_finally): Record outer_non_cleanup
1652 (lower_catch): Likewise.
1653 (lower_eh_filter): Likewise.
1654 (lower_eh_must_not_throw): Likewise.
1655 (lower_cleanup): Likewise.
1657 2020-01-20 Richard Biener <rguenther@suse.de>
1659 PR tree-optimization/93094
1660 * tree-vectorizer.h (vect_loop_versioning): Adjust.
1661 (vect_transform_loop): Likewise.
1662 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
1663 loop_vectorized_call to vect_transform_loop.
1664 * tree-vect-loop.c (vect_transform_loop): Pass down
1665 loop_vectorized_call to vect_loop_versioning.
1666 * tree-vect-loop-manip.c (vect_loop_versioning): Use
1667 the earlier discovered loop_vectorized_call.
1669 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
1671 * doc/contribute.texi: Update for SVN -> Git transition.
1672 * doc/install.texi: Likewise.
1674 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1677 * cgraph.c (cgraph_edge::make_speculative): Increase number of
1678 speculative targets.
1679 (verify_speculative_call): New function
1680 (cgraph_node::verify_node): Use it.
1681 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
1684 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1687 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
1688 (cgraph_edge::make_direct): Remove all indirect targets.
1689 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
1690 (cgraph_node::verify_node): Verify that only one call_stmt or
1691 lto_stmt_uid is set.
1692 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
1694 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
1695 (lto_output_ref): Simplify streaming of stmt.
1696 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
1698 2020-01-18 Tamar Christina <tamar.christina@arm.com>
1700 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
1701 Mark parameter unused.
1703 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
1705 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
1707 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
1709 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
1711 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
1713 * Makefile.in: Add coroutine-passes.o.
1714 * builtin-types.def (BT_CONST_SIZE): New.
1715 (BT_FN_BOOL_PTR): New.
1716 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
1717 * builtins.def (DEF_COROUTINE_BUILTIN): New.
1718 * coroutine-builtins.def: New file.
1719 * coroutine-passes.cc: New file.
1720 * function.h (struct GTY function): Add a bit to indicate that the
1721 function is a coroutine component.
1722 * internal-fn.c (expand_CO_FRAME): New.
1723 (expand_CO_YIELD): New.
1724 (expand_CO_SUSPN): New.
1725 (expand_CO_ACTOR): New.
1726 * internal-fn.def (CO_ACTOR): New.
1730 * passes.def: Add pass_coroutine_lower_builtins,
1731 pass_coroutine_early_expand_ifns.
1732 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
1733 (make_pass_coroutine_early_expand_ifns): New.
1734 * doc/invoke.texi: Document the fcoroutines command line
1737 2020-01-18 Jakub Jelinek <jakub@redhat.com>
1739 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
1742 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
1743 after checking the argument is a REG. Don't use REGNO (reg)
1744 again to set last_regno, reuse regno variable instead.
1746 2020-01-17 David Malcolm <dmalcolm@redhat.com>
1748 * doc/analyzer.texi (Limitations): Add note about NaN.
1750 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1751 Sudakshina Das <sudi.das@arm.com>
1753 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
1754 and valid immediate.
1755 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
1756 (lshrdi3): Generate thumb2_lsrl for valid immediates.
1757 * config/arm/constraints.md (Pg): New.
1758 * config/arm/predicates.md (long_shift_imm): New.
1759 (arm_reg_or_long_shift_imm): Likewise.
1760 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
1761 (thumb2_lsll): Likewise.
1764 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1765 Sudakshina Das <sudi.das@arm.com>
1767 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
1768 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
1769 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
1770 register pairs for doubleword quantities for ARMv8.1M-Mainline.
1771 * config/arm/thumb2.md (thumb2_asrl): New.
1772 (thumb2_lsll): Likewise.
1774 2020-01-17 Jakub Jelinek <jakub@redhat.com>
1776 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
1779 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
1781 * gdbinit.in (help-gcc-hooks): New command.
1782 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
1783 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
1786 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1788 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
1789 correct target macro.
1791 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1793 * config/aarch64/aarch64-protos.h
1794 (aarch64_sve_ld1ro_operand_p): New.
1795 * config/aarch64/aarch64-sve-builtins-base.cc
1796 (class load_replicate): New.
1797 (class svld1ro_impl): New.
1798 (class svld1rq_impl): Change to inherit from load_replicate.
1799 (svld1ro): New sve intrinsic function base.
1800 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
1801 New DEF_SVE_FUNCTION.
1802 * config/aarch64/aarch64-sve-builtins-base.h
1803 (svld1ro): New decl.
1804 * config/aarch64/aarch64-sve-builtins.cc
1805 (function_expander::add_mem_operand): Modify assert to allow
1807 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
1809 * config/aarch64/aarch64.c
1810 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
1811 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
1812 (aarch64_sve_ld1ro_operand_p): New.
1813 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
1814 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
1815 * config/aarch64/predicates.md
1816 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
1818 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1820 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
1821 Introduce this ACLE specified predefined macro.
1822 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
1823 (fp): Disabling this disables f64mm.
1824 (simd): Disabling this disables f64mm.
1825 (fp16): Disabling this disables f64mm.
1826 (sve): Disabling this disables f64mm.
1827 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
1828 (AARCH64_ISA_F64MM): New.
1829 (TARGET_F64MM): New.
1830 * doc/invoke.texi (f64mm): Document new option.
1832 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
1834 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
1835 (neoversen1_tunings): Likewise.
1837 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
1840 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
1841 Add assert to ensure prolog has been emitted.
1842 (aarch64_split_atomic_op): Likewise.
1843 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
1844 Use epilogue_completed rather than reload_completed.
1845 (aarch64_atomic_exchange<mode>): Likewise.
1846 (aarch64_atomic_<atomic_optab><mode>): Likewise.
1847 (atomic_nand<mode>): Likewise.
1848 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
1849 (atomic_fetch_nand<mode>): Likewise.
1850 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
1851 (atomic_nand_fetch<mode>): Likewise.
1853 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
1856 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
1858 (REVERSE_CONDITION): Delete.
1859 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
1860 (CCFP_CCFPE): Likewise.
1861 (e): New mode attribute.
1862 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
1863 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
1864 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
1865 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
1866 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
1867 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
1868 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
1869 name of generator from gen_ccmpdi to gen_ccmpccdi.
1870 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
1871 the previous comparison but aren't able to, use the new ccmp_rev
1874 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
1876 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
1877 than testing directly for INTEGER_CST.
1878 (gimplify_target_expr, gimplify_omp_depend): Likewise.
1880 2020-01-17 Jakub Jelinek <jakub@redhat.com>
1882 PR tree-optimization/93292
1883 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
1884 get_vectype_for_scalar_type returns NULL.
1886 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
1888 * params.opt (-param=max-predicted-iterations): Increase range from 0.
1889 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
1891 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
1893 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
1895 * params.opt: (max-predicted-iterations): Set bounds.
1896 * predict.c (real_almost_one, real_br_prob_base,
1897 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
1898 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
1899 probabilities; do not truncate to reg_br_prob_bases.
1900 (estimate_loops_at_level): Pass max_cyclic_prob.
1901 (estimate_loops): Compute max_cyclic_prob.
1902 (estimate_bb_frequencies): Do not initialize real_*; update calculation
1904 * profile-count.c (profile_probability::to_sreal): New.
1905 * profile-count.h (class sreal): Move up in file.
1906 (profile_probability::to_sreal): Declare.
1908 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1911 (arm_invalid_conversion): New function for target hook.
1912 (arm_invalid_unary_op): New function for target hook.
1913 (arm_invalid_binary_op): New function for target hook.
1915 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1917 * config.gcc: Add arm_bf16.h.
1918 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
1919 (arm_simd_builtin_std_type): Add BFmode.
1920 (arm_init_simd_builtin_types): Define element types for vector types.
1921 (arm_init_bf16_types): New function.
1922 (arm_init_builtins): Add arm_init_bf16_types function call.
1923 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
1924 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
1925 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
1926 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
1927 (arm_vector_mode_supported_p): Add V4BF, V8BF.
1928 (arm_mangle_type): Add __bf16.
1929 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
1930 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
1931 arm_bf16_ptr_type_node.
1932 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
1933 define_split between ARM registers.
1934 * config/arm/arm_bf16.h: New file.
1935 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
1936 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
1938 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
1939 * config/arm/vfp.md: Add BFmode to movhf patterns.
1941 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
1942 Andre Vieira <andre.simoesdiasvieira@arm.com>
1944 * config/arm/arm-cpus.in (mve, mve_float): New features.
1945 (dsp, mve, mve.fp): New options.
1946 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
1947 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
1948 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
1950 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1951 Thomas Preud'homme <thomas.preudhomme@arm.com>
1953 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
1955 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
1956 error for using -mcmse when targeting Armv8.1-M Mainline.
1958 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1959 Thomas Preud'homme <thomas.preudhomme@arm.com>
1961 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
1962 address in r4 when targeting Armv8.1-M Mainline.
1963 (nonsecure_call_value_internal): Likewise.
1964 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
1965 a register match_operand again. Emit BLXNS when targeting
1967 (nonsecure_call_value_reg_thumb2): Likewise.
1969 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1970 Thomas Preud'homme <thomas.preudhomme@arm.com>
1972 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
1973 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
1974 variable as true when floating-point ABI is not hard. Replace
1975 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
1976 Generate VLSTM and VLLDM instruction respectively before and
1977 after a function call to cmse_nonsecure_call function.
1978 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
1979 (VUNSPEC_VLLDM): Likewise.
1980 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
1981 (lazy_load_multiple_insn): Likewise.
1983 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1984 Thomas Preud'homme <thomas.preudhomme@arm.com>
1986 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
1987 (arm_emit_vfp_multi_reg_pop): Likewise.
1988 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
1989 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
1990 restore callee-saved VFP registers.
1992 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1993 Thomas Preud'homme <thomas.preudhomme@arm.com>
1995 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
1996 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
1997 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
1998 callee-saved GPRs as well as clear ip register before doing a nonsecure
1999 call then restore callee-saved GPRs after it when targeting
2001 (arm_reorg): Adapt to function rename.
2003 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2004 Thomas Preud'homme <thomas.preudhomme@arm.com>
2006 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2007 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2008 clear_vfp_multiple pattern based on a new vfp parameter.
2009 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2010 targeting Armv8.1-M Mainline.
2011 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2012 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2013 whether VFP registers are available before looking call_used_regs for a
2015 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2016 of prototype of clear_operation_p.
2017 (clear_vfp_multiple_operation): New predicate.
2018 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2019 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2021 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2022 Thomas Preud'homme <thomas.preudhomme@arm.com>
2024 * config/arm/arm-protos.h (clear_operation_p): Declare.
2025 * config/arm/arm.c (clear_operation_p): New function.
2026 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2027 targeting Armv8.1-M Mainline or successor.
2028 (output_return_instruction): Only output APSR register clearing if
2029 Armv8.1-M Mainline instructions not available.
2030 (thumb_exit): Likewise.
2031 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2032 * config/arm/thumb2.md (clear_apsr): New define_insn.
2033 (clear_multiple): Likewise.
2034 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2036 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2037 Thomas Preud'homme <thomas.preudhomme@arm.com>
2039 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2040 (use_return_insn): Also return false for Armv8.1-M Mainline.
2041 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2042 Mainline instructions are available.
2043 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2044 when targeting Armv8.1-M Mainline Security Extensions.
2045 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2046 Mainline entry function.
2047 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2048 targeting Armv8.1-M Mainline or successor.
2049 (arm_expand_epilogue): Fix indentation of caller-saved register
2050 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2052 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2053 (FP_SYSREGS): Likewise.
2054 (enum vfp_sysregs_encoding): Define enum.
2055 (fp_sysreg_names): Declare.
2056 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2057 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2058 (pop_fpsysreg_insn): Likewise.
2060 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2061 Thomas Preud'homme <thomas.preudhomme@arm.com>
2063 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2064 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2065 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2066 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2067 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2068 (ARMv8_1m_main): New feature group.
2069 (armv8.1-m.main): New architecture.
2070 * config/arm/arm-tables.opt: Regenerate.
2071 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2072 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2073 (arm_options_perform_arch_sanity_checks): Error out when targeting
2074 Armv8.1-M Mainline Security Extensions.
2075 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2077 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2079 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2080 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2081 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2082 aarch64_bfdot_laneq): New.
2083 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2084 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2085 vbfdotq_laneq_f32): New.
2086 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2087 VBFMLA_W, VBF): New.
2088 (isquadop): Add V4BF, V8BF.
2090 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2092 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2093 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2094 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2095 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2096 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2097 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2098 usdot_laneq, sudot_lane,sudot_laneq): New.
2099 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2100 (aarch64_<sur>dot_lane): New.
2101 * config/aarch64/arm_neon.h (vusdot_s32): New.
2103 (vusdot_lane_s32): New.
2104 (vsudot_lane_s32): New.
2105 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2106 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2108 2020-01-16 Martin Liska <mliska@suse.cz>
2110 * value-prof.c (dump_histogram_value): Fix
2111 obvious spacing issue.
2113 2020-01-16 Andrew Pinski <apinski@marvell.com>
2115 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2116 !storage_order_barrier_p.
2118 2020-01-16 Andrew Pinski <apinski@marvell.com>
2120 * sched-int.h (_dep): Add unused bit-field field for the padding.
2121 * sched-deps.c (init_dep_1): Init unused field.
2123 2020-01-16 Andrew Pinski <apinski@marvell.com>
2125 * optabs.h (create_expand_operand): Initialize target field also.
2127 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2129 PR tree-optimization/92429
2130 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2131 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2133 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2136 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2138 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2139 aarch64_sve_int_mode to each mode.
2141 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2143 * doc/analyzer.texi (Overview): Add note about
2144 -fdump-ipa-analyzer.
2146 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2148 PR tree-optimization/93231
2149 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2150 input_type is unsigned. Use tree_to_shwi for shift constant.
2151 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2152 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2153 inputs more efficiently.
2155 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2157 * config/i386/i386.md (*movsf_internal): Do not require
2158 SSE2 ISA for alternatives 14 and 15.
2160 2020-01-15 Richard Biener <rguenther@suse.de>
2163 * tree-eh.c (sink_clobbers): If we already visited the destination
2164 block do not defer insertion.
2165 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2166 the purpose of defered insertion.
2168 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2170 * BASE-VER: Bump to 10.0.1.
2172 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2174 PR tree-optimization/93247
2175 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2176 type of the stmt that we're going to vectorize.
2178 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2180 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2181 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2184 2020-01-15 Martin Liska <mliska@suse.cz>
2186 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2187 2 calls of streamer_read_hwi in a function call.
2189 2020-01-15 Richard Biener <rguenther@suse.de>
2191 * alias.c (record_alias_subset): Avoid redundant work when
2192 subset is already recorded.
2194 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2196 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2197 the analyzer options provide CWE identifiers.
2199 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2201 * tree-diagnostic-path.cc (path_summary::event_range::print):
2202 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2203 using get_pure_location.
2205 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2207 PR tree-optimization/93262
2208 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2209 perform head trimming only if the last argument is constant,
2210 either all ones, or larger or equal to head trim, in the latter
2211 case decrease the last argument by head_trim.
2213 PR tree-optimization/93249
2214 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2215 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2216 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2217 perform head trim unless we can prove there are no '\0' chars
2218 from the source among the first head_trim chars.
2220 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2222 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2224 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2227 * config/i386/sse.md
2228 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2229 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2230 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2231 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2232 just a single alternative instead of two, make operands 1 and 2
2235 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2238 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2241 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2243 * Makefile.in (lang_opt_files): Add analyzer.opt.
2244 (ANALYZER_OBJS): New.
2245 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2246 tristate.o and ANALYZER_OBJS.
2247 (TEXI_GCCINT_FILES): Add analyzer.texi.
2248 * common.opt (-fanalyzer): New driver option.
2249 * config.in: Regenerate.
2250 * configure: Regenerate.
2251 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2252 (gccdepdir): Also create depdir for "analyzer" subdir.
2253 * digraph.cc: New file.
2254 * digraph.h: New file.
2255 * doc/analyzer.texi: New file.
2256 * doc/gccint.texi ("Static Analyzer") New menu item.
2257 (analyzer.texi): Include it.
2258 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2259 ("Warning Options"): Add static analysis warnings to the list.
2260 (-Wno-analyzer-double-fclose): New option.
2261 (-Wno-analyzer-double-free): New option.
2262 (-Wno-analyzer-exposure-through-output-file): New option.
2263 (-Wno-analyzer-file-leak): New option.
2264 (-Wno-analyzer-free-of-non-heap): New option.
2265 (-Wno-analyzer-malloc-leak): New option.
2266 (-Wno-analyzer-possible-null-argument): New option.
2267 (-Wno-analyzer-possible-null-dereference): New option.
2268 (-Wno-analyzer-null-argument): New option.
2269 (-Wno-analyzer-null-dereference): New option.
2270 (-Wno-analyzer-stale-setjmp-buffer): New option.
2271 (-Wno-analyzer-tainted-array-index): New option.
2272 (-Wno-analyzer-use-after-free): New option.
2273 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2274 (-Wno-analyzer-use-of-uninitialized-value): New option.
2275 (-Wanalyzer-too-complex): New option.
2276 (-fanalyzer-call-summaries): New warning.
2277 (-fanalyzer-checker=): New warning.
2278 (-fanalyzer-fine-grained): New warning.
2279 (-fno-analyzer-state-merge): New warning.
2280 (-fno-analyzer-state-purge): New warning.
2281 (-fanalyzer-transitivity): New warning.
2282 (-fanalyzer-verbose-edges): New warning.
2283 (-fanalyzer-verbose-state-changes): New warning.
2284 (-fanalyzer-verbosity=): New warning.
2285 (-fdump-analyzer): New warning.
2286 (-fdump-analyzer-callgraph): New warning.
2287 (-fdump-analyzer-exploded-graph): New warning.
2288 (-fdump-analyzer-exploded-nodes): New warning.
2289 (-fdump-analyzer-exploded-nodes-2): New warning.
2290 (-fdump-analyzer-exploded-nodes-3): New warning.
2291 (-fdump-analyzer-supergraph): New warning.
2292 * doc/sourcebuild.texi (dg-require-dot): New.
2293 (dg-check-dot): New.
2294 * gdbinit.in (break-on-saved-diagnostic): New command.
2295 * graphviz.cc: New file.
2296 * graphviz.h: New file.
2297 * ordered-hash-map-tests.cc: New file.
2298 * ordered-hash-map.h: New file.
2299 * passes.def (pass_analyzer): Add before
2300 pass_ipa_whole_program_visibility.
2301 * selftest-run-tests.c (selftest::run_tests): Call
2302 selftest::ordered_hash_map_tests_cc_tests.
2303 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2305 * shortest-paths.h: New file.
2306 * timevar.def (TV_ANALYZER): New timevar.
2307 (TV_ANALYZER_SUPERGRAPH): Likewise.
2308 (TV_ANALYZER_STATE_PURGE): Likewise.
2309 (TV_ANALYZER_PLAN): Likewise.
2310 (TV_ANALYZER_SCC): Likewise.
2311 (TV_ANALYZER_WORKLIST): Likewise.
2312 (TV_ANALYZER_DUMP): Likewise.
2313 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2314 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2315 * tree-pass.h (make_pass_analyzer): New decl.
2316 * tristate.cc: New file.
2317 * tristate.h: New file.
2319 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2322 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2323 alternatives 9 and 10.
2325 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2327 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2328 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2329 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2330 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2331 (selftest::hash_map_tests_c_tests): Call it.
2332 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2333 New static constant, using the value of = H::empty_zero_p.
2334 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2335 from default_hash_traits <Value>.
2336 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2338 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2339 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2340 calls to mark_empty with !Descriptor::empty_zero_p.
2341 (hash_table::empty_slow): Conditionalize the memset call with a
2342 check that Descriptor::empty_zero_p; otherwise, loop through the
2343 entries calling mark_empty on them.
2344 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2345 (pointer_hash::empty_zero_p): Likewise.
2346 (pair_hash::empty_zero_p): Likewise.
2347 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2349 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2350 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2351 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2352 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2353 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2354 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2355 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2356 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2358 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2361 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2363 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2364 fix typo on return value.
2366 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2369 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2371 (cgraph_edge::make_speculative): Add param for setting speculative_id
2373 (cgraph_edge::speculative_call_info): Update comments and find reference
2374 by speculative_id for multiple indirect targets.
2375 (cgraph_edge::resolve_speculation): Decrease the speculations
2376 for indirect edge, drop it's speculative if not direct target
2377 left. Update comments.
2378 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2379 (cgraph_node::dump): Print num_speculative_call_targets.
2380 (cgraph_node::verify_node): Don't report error if speculative
2381 edge not include statement.
2382 (cgraph_edge::num_speculative_call_targets_p): New function.
2383 * cgraph.h (int common_target_id): Remove.
2384 (int common_target_probability): Remove.
2385 (num_speculative_call_targets): New variable.
2386 (make_speculative): Add param for setting speculative_id.
2387 (cgraph_edge::num_speculative_call_targets_p): New declare.
2388 (target_prob): New variable.
2389 (speculative_id): New variable.
2390 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2391 call summaries for multiple speculative call targets.
2392 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2393 * ipa-profile.c (struct speculative_call_target): New struct.
2394 (class speculative_call_summary): New class.
2395 (class speculative_call_summaries): New class.
2396 (call_sums): New variable.
2397 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2398 (ipa_profile_write_edge_summary): New function.
2399 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2400 (ipa_profile_dump_all_summaries): New function.
2401 (ipa_profile_read_edge_summary): New function.
2402 (ipa_profile_read_summary_section): New function.
2403 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2404 (ipa_profile): Generate num_speculative_call_targets from
2406 * ipa-ref.h (speculative_id): New variable.
2407 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2408 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2409 common_target_probability. Stream out speculative_id and
2410 num_speculative_call_targets.
2411 (input_edge): Likewise.
2412 * predict.c (dump_prediction): Remove edges count assert to be
2414 * symtab.c (symtab_node::create_reference): Init speculative_id.
2415 (symtab_node::clone_references): Clone speculative_id.
2416 (symtab_node::clone_referring): Clone speculative_id.
2417 (symtab_node::clone_reference): Clone speculative_id.
2418 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2419 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2420 if indirect call contains multiple speculative targets.
2421 * value-prof.h (check_ic_target): Remove.
2422 * value-prof.c (gimple_value_profile_transformations):
2423 Use void function gimple_ic_transform.
2424 * value-prof.c (gimple_ic_transform): Handle topn case.
2425 Fix comment typos. Change it to a void function.
2427 2020-01-13 Andrew Pinski <apinski@marvell.com>
2429 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2430 (octeontx2t98): New define.
2431 (octeontx2t96): New define.
2432 (octeontx2t93): New define.
2433 (octeontx2f95): New define.
2434 (octeontx2f95n): New define.
2435 (octeontx2f95mm): New define.
2436 * config/aarch64/aarch64-tune.md: Regenerate.
2437 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2439 2020-01-13 Jason Merrill <jason@redhat.com>
2441 PR c++/33799 - destroy return value if local cleanup throws.
2442 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2444 2020-01-13 Martin Liska <mliska@suse.cz>
2446 * ipa-cp.c (get_max_overall_size): Use newly
2447 renamed param param_ipa_cp_unit_growth.
2448 * params.opt: Remove legacy param name.
2450 2020-01-13 Martin Sebor <msebor@redhat.com>
2452 PR tree-optimization/93213
2453 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2454 stores to be eliminated.
2456 2020-01-13 Martin Liska <mliska@suse.cz>
2458 * opts.c (print_help): Do not print CL_PARAM
2459 and CL_WARNING for CL_OPTIMIZATION.
2461 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2464 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2465 depending on optimization settings.
2467 2020-01-13 Jakub Jelinek <jakub@redhat.com>
2469 PR tree-optimization/90838
2470 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2471 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2472 argument rather than to initialize temporary for targets that
2473 don't use the mode argument at all. Initialize ctzval to avoid
2476 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
2478 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2479 * tree-core.h: Document it.
2480 * gimplify.c (gimplify_omp_workshare): Set it.
2481 * omp-low.c (lower_omp_target): Use it.
2482 * tree-pretty-print.c (dump_omp_clause): Print it.
2484 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2485 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2487 2020-01-10 David Malcolm <dmalcolm@redhat.com>
2489 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2490 * common.opt (fdiagnostics-path-format=): New option.
2491 (diagnostic_path_format): New enum.
2492 (fdiagnostics-show-path-depths): New option.
2493 * coretypes.h (diagnostic_event_id_t): New forward decl.
2494 * diagnostic-color.c (color_dict): Add "path".
2495 * diagnostic-event-id.h: New file.
2496 * diagnostic-format-json.cc (json_from_expanded_location): Make
2498 (json_end_diagnostic): Call context->make_json_for_path if it
2499 exists and the diagnostic has a path.
2500 (diagnostic_output_format_init): Clear context->print_path.
2501 * diagnostic-path.h: New file.
2502 * diagnostic-show-locus.c (colorizer::set_range): Special-case
2503 when printing a run of events in a diagnostic_path so that they
2504 all get the same color.
2505 (layout::m_diagnostic_path_p): New field.
2506 (layout::layout): Initialize it.
2507 (layout::print_any_labels): Don't colorize the label text for an
2508 event in a diagnostic_path.
2509 (gcc_rich_location::add_location_if_nearby): Add
2510 "restrict_to_current_line_spans" and "label" params. Pass the
2511 former to layout.maybe_add_location_range; pass the latter
2512 when calling add_range.
2513 * diagnostic.c: Include "diagnostic-path.h".
2514 (diagnostic_initialize): Initialize context->path_format and
2515 context->show_path_depths.
2516 (diagnostic_show_any_path): New function.
2517 (diagnostic_path::interprocedural_p): New function.
2518 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2519 (simple_diagnostic_path::num_events): New function.
2520 (simple_diagnostic_path::get_event): New function.
2521 (simple_diagnostic_path::add_event): New function.
2522 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2523 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
2524 (debug): New overload taking a diagnostic_path *.
2525 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
2526 * diagnostic.h (enum diagnostic_path_format): New enum.
2527 (json::value): New forward decl.
2528 (diagnostic_context::path_format): New field.
2529 (diagnostic_context::show_path_depths): New field.
2530 (diagnostic_context::print_path): New callback field.
2531 (diagnostic_context::make_json_for_path): New callback field.
2532 (diagnostic_show_any_path): New decl.
2533 (json_from_expanded_location): New decl.
2534 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
2535 (-fdiagnostics-show-path-depths): New option.
2536 (-fdiagnostics-color): Add "path" to description of default
2537 GCC_COLORS; describe it.
2538 (-fdiagnostics-format=json): Document how diagnostic paths are
2539 represented in the JSON output format.
2540 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2541 Add optional params "restrict_to_current_line_spans" and "label".
2542 * opts.c (common_handle_option): Handle
2543 OPT_fdiagnostics_path_format_ and
2544 OPT_fdiagnostics_show_path_depths.
2545 * pretty-print.c: Include "diagnostic-event-id.h".
2546 (pp_format): Implement "%@" format code for printing
2547 diagnostic_event_id_t *.
2548 (selftest::test_pp_format): Add tests for "%@".
2549 * selftest-run-tests.c (selftest::run_tests): Call
2550 selftest::tree_diagnostic_path_cc_tests.
2551 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
2552 * toplev.c (general_init): Initialize global_dc->path_format and
2553 global_dc->show_path_depths.
2554 * tree-diagnostic-path.cc: New file.
2555 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
2556 non-static. Drop "diagnostic" param in favor of storing the
2557 original value of "where" and re-using it.
2558 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
2559 maybe_unwind_expanded_macro_loc.
2560 (tree_diagnostics_defaults): Initialize context->print_path and
2561 context->make_json_for_path.
2562 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
2564 (default_tree_make_json_for_path): New decl.
2565 (maybe_unwind_expanded_macro_loc): New decl.
2567 2020-01-10 Jakub Jelinek <jakub@redhat.com>
2569 PR tree-optimization/93210
2570 * fold-const.h (native_encode_initializer,
2571 can_native_interpret_type_p): Declare.
2572 * fold-const.c (native_encode_string): Fix up handling with off != -1,
2574 (native_encode_initializer): New function, moved from dwarf2out.c.
2575 Adjust to native_encode_expr compatible arguments, including dry-run
2576 and partial extraction modes. Don't handle STRING_CST.
2577 (can_native_interpret_type_p): No longer static.
2578 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
2579 offset / BITS_PER_UNIT fits into int and don't call it if
2580 can_native_interpret_type_p fails. If suboff is NULL and for
2581 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
2582 native_encode_initializer.
2583 (fold_const_aggregate_ref_1): Formatting fix.
2584 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
2585 (tree_add_const_value_attribute): Adjust caller.
2587 PR tree-optimization/90838
2588 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2589 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
2590 CTZ_DEFINED_VALUE_AT_ZERO.
2592 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
2595 * lra-constraints.c (match_reload): Permit input operands have the
2596 same mode as output while other input operands have a different
2599 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
2601 PR tree-optimization/90838
2602 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
2603 (check_ctz_string): Likewise.
2604 (optimize_count_trailing_zeroes): Likewise.
2605 (simplify_count_trailing_zeroes): Likewise.
2606 (pass_forwprop::execute): Try ctz simplification.
2607 * match.pd: Add matching for ctz idioms.
2609 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2611 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
2613 (aarch64_invalid_unary_op): New function for target hook.
2614 (aarch64_invalid_binary_op): New function for target hook.
2616 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2618 * config.gcc: Add arm_bf16.h.
2619 * config/aarch64/aarch64-builtins.c
2620 (aarch64_simd_builtin_std_type): Add BFmode.
2621 (aarch64_init_simd_builtin_types): Define element types for vector
2623 (aarch64_init_bf16_types): New function.
2624 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
2625 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
2627 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
2628 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
2630 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
2631 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
2632 * config/aarch64/aarch64.c
2633 (aarch64_classify_vector_mode): Add support for BF types.
2634 (aarch64_gimplify_va_arg_expr): Add support for BF types.
2635 (aarch64_vq_mode): Add support for BF types.
2636 (aarch64_simd_container_mode): Add support for BF types.
2637 (aarch64_mangle_type): Add support for BF scalar type.
2638 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
2639 * config/aarch64/arm_bf16.h: New file.
2640 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2641 * config/aarch64/iterators.md: Add BF types to mode attributes.
2642 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
2644 2020-01-10 Jason Merrill <jason@redhat.com>
2646 PR c++/93173 - incorrect tree sharing.
2647 * gimplify.c (copy_if_shared): No longer static.
2648 * gimplify.h: Declare it.
2650 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2652 * doc/invoke.texi (-msve-vector-bits=): Document that
2653 -msve-vector-bits=128 now generates VL-specific code for
2654 little-endian targets.
2655 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
2656 build_vector_type_for_mode to construct the data vector types.
2657 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
2658 VL-specific code for -msve-vector-bits=128 on little-endian targets.
2659 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
2660 for 128-bit vectors.
2662 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2664 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
2667 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2669 * config/aarch64/aarch64-builtins.c
2670 (aarch64_builtin_vectorized_function): Check for specific vector modes,
2671 rather than checking the number of elements and the element mode.
2673 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2675 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
2676 get_related_vectype_for_scalar_type rather than build_vector_type
2677 to create the index type for a conditional reduction.
2679 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2681 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
2682 for any type of gather or scatter, including strided accesses.
2684 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2686 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
2689 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2691 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
2693 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
2694 parameter and its use to reset DR_OFFSET's.
2695 (vect_transform_loop): Remove orig_drs_init argument.
2696 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
2697 member of dr_vec_info rather than the offset of the associated
2698 data_reference's innermost_loop_behavior.
2699 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
2700 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
2701 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
2702 get_dr_vinfo_offset.
2703 (vectorizable_store): Likewise.
2704 (vectorizable_load): Likewise.
2706 2020-01-10 Richard Biener <rguenther@suse.de>
2708 * gimple-ssa-store-merging
2709 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
2711 2020-01-10 Martin Liska <mliska@suse.cz>
2714 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
2715 encapsulation that was there before r280040.
2717 2020-01-10 Richard Biener <rguenther@suse.de>
2720 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
2721 sequences to avoid walking them again for secondary opportunities.
2722 (pass_lower_eh_dispatch::execute): Instead actually insert
2725 2020-01-10 Richard Biener <rguenther@suse.de>
2728 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
2729 (cleanup_all_empty_eh): Walk landing pads in reverse order to
2730 avoid quadraticness.
2732 2020-01-10 Martin Jambor <mjambor@suse.cz>
2734 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
2735 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
2736 to get param_ipa_sra_max_replacements.
2737 (param_splitting_across_edge): Pass the caller to
2738 pull_accesses_from_callee.
2740 2020-01-10 Martin Jambor <mjambor@suse.cz>
2742 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
2743 * ipa-cp.c (max_new_size): Removed.
2744 (orig_overall_size): New variable.
2745 (get_max_overall_size): New function.
2746 (estimate_local_effects): Use it. Adjust dump.
2747 (decide_about_value): Likewise.
2748 (ipcp_propagate_stage): Do not calculate max_new_size, just store
2749 orig_overall_size. Adjust dump.
2750 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
2752 2020-01-10 Martin Jambor <mjambor@suse.cz>
2754 * params.opt (param_ipa_max_agg_items): Mark as Optimization
2755 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
2756 instead of param_ipa_max_agg_items.
2757 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
2758 optimization info for the callee.
2760 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
2762 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
2763 markers if debug_inline_points is false.
2765 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2767 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
2769 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
2770 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
2771 aarch64-sve-builtins-sve2.h.
2772 (aarch64-sve-builtins-sve2.o): New rule.
2773 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
2774 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
2775 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
2776 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
2777 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
2778 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
2780 * config/aarch64/aarch64-sve.md: Update comments with SVE2
2781 instructions that are handled here.
2782 (@cond_asrd<mode>): Generalize to...
2783 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
2784 (*cond_asrd<mode>_2): Generalize to...
2785 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
2786 (*cond_asrd<mode>_z): Generalize to...
2787 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
2788 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
2789 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
2790 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
2791 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
2793 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2794 (@aarch64_scatter_stnt<mode>): Likewise.
2795 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2796 (@aarch64_mul_lane_<mode>): Likewise.
2797 (@aarch64_sve_suqadd<mode>_const): Likewise.
2798 (*<sur>h<addsub><mode>): Generalize to...
2799 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
2801 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
2802 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
2803 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
2804 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
2805 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
2806 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
2807 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
2808 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
2809 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
2810 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
2811 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
2812 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
2813 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
2814 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
2815 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
2816 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
2817 (@aarch64_sve2_xar<mode>): Likewise.
2818 (@aarch64_sve2_bcax<mode>): Likewise.
2819 (*aarch64_sve2_eor3<mode>): Rename to...
2820 (@aarch64_sve2_eor3<mode>): ...this.
2821 (@aarch64_sve2_bsl<mode>): New expander.
2822 (@aarch64_sve2_nbsl<mode>): Likewise.
2823 (@aarch64_sve2_bsl1n<mode>): Likewise.
2824 (@aarch64_sve2_bsl2n<mode>): Likewise.
2825 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
2826 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
2827 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
2828 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
2829 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
2830 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
2831 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
2832 (<su>mull<bt><Vwide>): Generalize to...
2833 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
2835 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
2836 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
2837 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
2838 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2839 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
2840 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2841 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
2842 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2843 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
2844 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2845 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
2846 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
2847 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
2848 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
2849 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
2850 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
2851 (<SHRNB:r>shrnb<mode>): Generalize to...
2852 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
2854 (<SHRNT:r>shrnt<mode>): Generalize to...
2855 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
2857 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
2858 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
2859 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
2860 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
2861 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
2862 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
2863 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
2864 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
2865 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
2866 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
2867 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
2868 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
2869 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
2870 (@aarch64_sve2_cvtnt<mode>): Likewise.
2871 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
2872 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
2873 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
2874 (@aarch64_sve2_cvtxnt<mode>): Likewise.
2875 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
2876 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
2877 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
2878 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
2879 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
2880 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
2881 (@aarch64_sve2_pmul<mode>): Likewise.
2882 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
2883 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
2884 (@aarch64_sve2_tbl2<mode>): Likewise.
2885 (@aarch64_sve2_tbx<mode>): Likewise.
2886 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
2887 (@aarch64_sve2_histcnt<mode>): Likewise.
2888 (@aarch64_sve2_histseg<mode>): Likewise.
2889 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
2890 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
2891 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
2892 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
2893 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
2894 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
2895 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
2896 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
2897 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
2898 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
2899 (SVE2_PMULL_PAIR_I): New mode iterators.
2900 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
2901 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
2902 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
2903 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
2904 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
2905 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
2906 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
2907 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
2908 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
2909 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
2910 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
2911 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
2912 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
2913 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
2914 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
2915 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
2916 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
2917 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
2918 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
2919 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
2920 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
2921 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
2922 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
2923 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
2924 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
2925 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
2926 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
2927 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
2928 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
2929 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
2930 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
2932 (VNARROW, Ventype): New mode attributes.
2933 (Vewtype): Handle VNx2DI. Fix typo in comment.
2934 (VDOUBLE): New mode attribute.
2935 (sve_lane_con): Handle VNx8HI.
2936 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
2937 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
2938 (sve_int_op, sve_int_op_rev): Handle the above codes.
2939 (sve_pred_int_rhs2_operand): Likewise.
2940 (MULLBT, SHRNB, SHRNT): Delete.
2941 (SVE_INT_SHIFT_IMM): New int iterator.
2942 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
2943 and UNSPEC_WHILEHS for TARGET_SVE2.
2944 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
2945 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
2946 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
2947 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
2948 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
2949 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
2950 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
2951 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
2952 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
2953 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
2954 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
2955 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
2956 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
2957 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
2958 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
2959 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
2960 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
2961 (optab): Handle the new unspecs.
2962 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
2964 (lr): Handle the new unspecs.
2966 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
2967 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
2968 (sve_int_qsub_op): New int attributes.
2969 (sve_fp_op, rot): Handle the new unspecs.
2970 * config/aarch64/aarch64-sve-builtins.h
2971 (function_resolver::require_matching_pointer_type): Declare.
2972 (function_resolver::resolve_unary): Add an optional boolean argument.
2973 (function_resolver::finish_opt_n_resolution): Add an optional
2974 type_suffix_index argument.
2975 (gimple_folder::redirect_call): Declare.
2976 (gimple_expander::prepare_gather_address_operands): Add an optional
2978 * config/aarch64/aarch64-sve-builtins.cc: Include
2979 aarch64-sve-builtins-sve2.h.
2980 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
2981 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
2982 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
2983 (TYPES_hsd_integer): Use TYPES_hsd_signed.
2984 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
2985 (TYPES_s_unsigned): Likewise.
2986 (TYPES_s_integer): Use TYPES_s_unsigned.
2987 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
2988 (TYPES_sd_integer): Use them.
2989 (TYPES_d_unsigned): New macro.
2990 (TYPES_d_integer): Use it.
2991 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
2992 (TYPES_cvt_narrow): Likewise.
2993 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
2994 (preds_mx): New variable.
2995 (function_builder::add_overloaded_function): Allow the new feature
2996 set to be more restrictive than the original one.
2997 (function_resolver::infer_pointer_type): Remove qualifiers from
2998 the pointer type before printing it.
2999 (function_resolver::require_matching_pointer_type): New function.
3000 (function_resolver::resolve_sv_displacement): Handle functions
3001 that don't support 32-bit vector indices or svint32_t vector offsets.
3002 (function_resolver::finish_opt_n_resolution): Take the inferred type
3003 as a separate argument.
3004 (function_resolver::resolve_unary): Optionally treat all forms in
3005 the same way as normal merging functions.
3006 (gimple_folder::redirect_call): New function.
3007 (function_expander::prepare_gather_address_operands): Add an argument
3008 that says whether scaled forms are available. If they aren't,
3009 handle scaling of vector indices and don't add the extension and
3011 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3012 fall back to using cond_* instead.
3013 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3014 Split out the member variables into...
3015 (rtx_code_function_base): ...this new base class.
3016 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3017 (unspec_based_function): Split out the member variables into...
3018 (unspec_based_function_base): ...this new base class.
3019 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3020 (unspec_based_function_exact_insn): New class.
3021 (unspec_based_add_function, unspec_based_add_lane_function)
3022 (unspec_based_lane_function, unspec_based_pred_function)
3023 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3024 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3025 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3027 (unspec_based_fused_function): New class.
3028 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3029 (unspec_based_fused_lane_function): New class.
3030 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3032 (CODE_FOR_MODE1): New macro.
3033 (fixed_insn_function): New class.
3034 (while_comparison): Likewise.
3035 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3036 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3037 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3038 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3039 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3040 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3041 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3042 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3043 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3044 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3045 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3046 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3047 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3048 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3049 Also add an initial argument for unary_convert_narrowt, regardless
3050 of the predication type.
3051 (build_32_64): Allow loads and stores to specify MODE_none.
3052 (build_sv_index64, build_sv_uint_offset): New functions.
3053 (long_type_suffix): New function.
3054 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3055 (binary_imm_long_base, load_gather_sv_base): Likewise.
3056 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3057 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3058 (unary_narrowb_base, unary_narrowt_base): Likewise.
3059 (binary_long_lane_def, binary_long_lane): New shape.
3060 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3061 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3062 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3063 (binary_to_uint_def, binary_to_uint): Likewise.
3064 (binary_wide_def, binary_wide): Likewise.
3065 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3066 (compare_def, compare): Likewise.
3067 (compare_ptr_def, compare_ptr): Likewise.
3068 (load_ext_gather_index_restricted_def,
3069 load_ext_gather_index_restricted): Likewise.
3070 (load_ext_gather_offset_restricted_def,
3071 load_ext_gather_offset_restricted): Likewise.
3072 (load_gather_sv_def): Inherit from load_gather_sv_base.
3073 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3074 (shift_left_imm_def, shift_left_imm): Likewise.
3075 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3076 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3077 (store_scatter_index_restricted_def,
3078 store_scatter_index_restricted): Likewise.
3079 (store_scatter_offset_restricted_def,
3080 store_scatter_offset_restricted): Likewise.
3081 (tbl_tuple_def, tbl_tuple): Likewise.
3082 (ternary_long_lane_def, ternary_long_lane): Likewise.
3083 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3084 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3085 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3086 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3087 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3088 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3089 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3090 (ternary_uint_def, ternary_uint): Likewise.
3091 (unary_convert): Fix typo in comment.
3092 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3093 (unary_long_def, unary_long): Likewise.
3094 (unary_narrowb_def, unary_narrowb): Likewise.
3095 (unary_narrowt_def, unary_narrowt): Likewise.
3096 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3097 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3098 (unary_to_int_def, unary_to_int): Likewise.
3099 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3100 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3101 (svasrd_impl): Delete.
3102 (svcadd_impl::expand): Handle integer operations too.
3103 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3104 new functions to derive the unspec numbers.
3105 (svmla_svmls_lane_impl): Replace with...
3106 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3107 integer operations too.
3108 (svwhile_impl): Rename to...
3109 (svwhilelx_impl): ...this and inherit from while_comparison.
3110 (svasrd): Use unspec_based_function.
3111 (svmla_lane): Use svmla_lane_impl.
3112 (svmls_lane): Use svmls_lane_impl.
3113 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3114 (svwhilele, svwhilelt): Use svwhilelx_impl.
3115 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3116 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3117 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3118 * config/aarch64/aarch64-sve-builtins.def: Include
3119 aarch64-sve-builtins-sve2.def.
3121 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3123 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3124 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3125 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3126 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3127 immediates as well as vector ones.
3128 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3129 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3130 (aarch64_sve_qsub_immediate): Update calls accordingly.
3132 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3134 * config/aarch64/aarch64-sve2.md: Add banner comments.
3135 (<su>mulh<r>s<mode>3): Move further up file.
3136 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3137 (*aarch64_sve2_sra<mode>): Move further down file.
3138 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3140 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3142 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3144 (while_optab_cmp): Handle them.
3145 * config/aarch64/aarch64-sve.md
3146 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3147 and add a "@" marker.
3148 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3149 instead of gen_aarch64_sve2_while_ptest.
3150 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3152 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3154 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3155 (UNSPEC_WHILELE): ...this.
3156 (UNSPEC_WHILE_LO): Rename to...
3157 (UNSPEC_WHILELO): ...this.
3158 (UNSPEC_WHILE_LS): Rename to...
3159 (UNSPEC_WHILELS): ...this.
3160 (UNSPEC_WHILE_LT): Rename to...
3161 (UNSPEC_WHILELT): ...this.
3162 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3163 (cmp_op, while_optab_cmp): Likewise.
3164 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3165 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3166 (svwhilelt): Likewise.
3168 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3170 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3171 (unary_to_uint): Define.
3172 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3173 (unary_count): Rename to...
3174 (unary_to_uint_def, unary_to_uint): ...this.
3175 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3177 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3179 * config/aarch64/aarch64-sve-builtins-functions.h
3180 (code_for_mode_function): New class.
3181 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3182 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3183 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3184 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3185 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3187 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3189 * config/aarch64/iterators.md (addsub): New code attribute.
3190 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3192 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3193 in the asm string and attributes. Fix indentation.
3194 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3196 (@aarch64_sve_<optab><mode>): ...this.
3197 * config/aarch64/aarch64-sve-builtins.h
3198 (function_expander::expand_signed_unpred_op): Delete.
3199 * config/aarch64/aarch64-sve-builtins.cc
3200 (function_expander::expand_signed_unpred_op): Likewise.
3201 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3202 try using code_for_aarch64_sve instead.
3203 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3204 (svqsub_impl): Likewise.
3205 (svqadd, svqsub): Use rtx_code_function instead.
3207 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3209 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3210 (HADDSUB, sur, addsub): Remove them.
3212 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3214 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3215 functions the same way, rather than singling out those that
3216 aren't mapped directly to optabs.
3218 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3220 * target.def (compatible_vector_types_p): New target hook.
3221 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3222 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3223 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3224 * doc/tm.texi: Regenerate.
3225 * gimple-expr.c: Include target.h.
3226 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3227 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3229 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3230 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3231 Use the original predicate if it already has a suitable type.
3233 2020-01-09 Martin Jambor <mjambor@suse.cz>
3235 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3236 resolve_speculation and redirect_call_stmt_to_callee static. Change
3237 return type of set_call_stmt to cgraph_edge *.
3238 * auto-profile.c (afdo_indirect_call): Adjust call to
3239 redirect_call_stmt_to_callee.
3240 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3241 make the this pointer explicit, adjust self-recursive calls and the
3242 call top make_direct. Return the resulting edge.
3243 (cgraph_edge::remove): Make this pointer explicit.
3244 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3245 (cgraph_edge::make_direct): Likewise, adjust call to
3246 resolve_speculation.
3247 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3248 call to set_call_stmt.
3249 (cgraph_update_edges_for_call_stmt_node): Update call to
3250 set_call_stmt and remove.
3251 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3252 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3253 (cgraph_node::create_edge_including_clones): Moved "first" definition
3254 of edge to the block where it was used. Adjusted calls to
3256 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3257 cgraph_edge::remove.
3258 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3259 make_direct and redirect_call_stmt_to_callee.
3260 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3261 resolve_speculation and make_direct.
3262 * ipa-inline-transform.c (inline_transform): Adjust call to
3263 redirect_call_stmt_to_callee.
3264 (check_speculations_1):: Adjust call to resolve_speculation.
3265 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3266 resolve-speculation.
3267 (inline_small_functions): Adjust call to resolve_speculation.
3268 (ipa_inline): Likewise.
3269 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3271 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3272 safe with regards to edge removal, adjust calls to
3273 redirect_call_stmt_to_callee.
3274 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3275 and redirect_call_stmt_to_callee.
3276 * multiple_target.c (create_dispatcher_calls): Adjust call to
3277 redirect_call_stmt_to_callee
3278 (redirect_to_specific_clone): Likewise.
3279 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3280 Adjust calls to cgraph_edge::remove.
3281 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3282 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3283 (expand_call_inline): Adjust call to cgraph_edge::remove.
3285 2020-01-09 Martin Liska <mliska@suse.cz>
3287 * params.opt: Set Optimization for
3288 param_max_speculative_devirt_maydefs.
3290 2020-01-09 Martin Sebor <msebor@redhat.com>
3294 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3296 2020-01-09 Martin Liska <mliska@suse.cz>
3298 * auto-profile.c (auto_profile): Use opt_for_fn
3300 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3301 (propagate_vals_across_arith_jfunc): Likewise.
3302 (hint_time_bonus): Likewise.
3303 (incorporate_penalties): Likewise.
3304 (good_cloning_opportunity_p): Likewise.
3305 (perform_estimation_of_a_value): Likewise.
3306 (estimate_local_effects): Likewise.
3307 (ipcp_propagate_stage): Likewise.
3308 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3309 (set_switch_stmt_execution_predicate): Likewise.
3310 (analyze_function_body): Likewise.
3311 * ipa-inline-analysis.c (offline_size): Likewise.
3312 * ipa-inline.c (early_inliner): Likewise.
3313 * ipa-prop.c (ipa_analyze_node): Likewise.
3314 (ipcp_transform_function): Likewise.
3315 * ipa-sra.c (process_scan_results): Likewise.
3316 (ipa_sra_summarize_function): Likewise.
3317 * params.opt: Rename ipcp-unit-growth to
3318 ipa-cp-unit-growth. Add Optimization for various
3319 IPA-related parameters.
3321 2020-01-09 Richard Biener <rguenther@suse.de>
3324 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3326 2020-01-09 Richard Biener <rguenther@suse.de>
3328 PR tree-optimization/93040
3329 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3331 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3333 * common/config/avr/avr-common.c (avr_option_optimization_table)
3334 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3336 2020-01-09 Martin Liska <mliska@suse.cz>
3338 * cgraphclones.c (symbol_table::materialize_all_clones):
3339 Use cgraph_node::dump_name.
3341 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3344 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3345 output_operand_lossage instead of gcc_unreachable.
3346 * doc/md.texi (riscv f constraint): Fix typo.
3349 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3350 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3351 CONST_SCALAR_INT_P instead of CONST_INT_P.
3352 (*subv<mode>4_1): Rename to ...
3353 (subv<mode>4_1): ... this.
3354 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3355 define_insn_and_split patterns.
3356 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3359 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3361 * vec.c (class selftest::count_dtor): New class.
3362 (selftest::test_auto_delete_vec): New test.
3363 (selftest::vec_c_tests): Call it.
3364 * vec.h (class auto_delete_vec): New class template.
3365 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3367 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3369 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3371 2020-01-08 Jim Wilson <jimw@sifive.com>
3373 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3374 use of TLS_MODEL_LOCAL_EXEC when not pic.
3376 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3378 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3381 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3384 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3385 *stack_protect_set_3 peephole2): Also check that the second
3386 insns source is general_operand.
3389 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3390 predicate for output operand instead of register_operand.
3391 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3392 memory destination and non-memory operands[2].
3394 2020-01-08 Martin Liska <mliska@suse.cz>
3396 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3397 ::dump_asm_name instead of (::name or ::asm_name).
3398 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3399 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3400 (analyze_functions): Likewise.
3401 (expand_all_functions): Likewise.
3402 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3403 (propagate_bits_across_jump_function): Likewise.
3404 (dump_profile_updates): Likewise.
3405 (ipcp_store_bits_results): Likewise.
3406 (ipcp_store_vr_results): Likewise.
3407 * ipa-devirt.c (dump_targets): Likewise.
3408 * ipa-fnsummary.c (analyze_function_body): Likewise.
3409 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3410 (process_hsa_functions): Likewise.
3411 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3412 (set_alias_uids): Likewise.
3413 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3414 * ipa-inline.c (recursive_inlining): Likewise.
3415 (inline_to_all_callers_1): Likewise.
3416 (ipa_inline): Likewise.
3417 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3418 (ipa_propagate_frequency): Likewise.
3419 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3420 (remove_described_reference): Likewise.
3421 * ipa-pure-const.c (worse_state): Likewise.
3422 (check_retval_uses): Likewise.
3423 (analyze_function): Likewise.
3424 (propagate_pure_const): Likewise.
3425 (propagate_nothrow): Likewise.
3426 (dump_malloc_lattice): Likewise.
3427 (propagate_malloc): Likewise.
3428 (pass_local_pure_const::execute): Likewise.
3429 * ipa-visibility.c (optimize_weakref): Likewise.
3430 (function_and_variable_visibility): Likewise.
3431 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3432 (ipa_discover_variable_flags): Likewise.
3433 * lto-streamer-out.c (output_function): Likewise.
3434 (output_constructor): Likewise.
3435 * tree-inline.c (copy_bb): Likewise.
3436 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3437 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3439 2020-01-08 Richard Biener <rguenther@suse.de>
3442 * tree-eh.c (sink_clobbers): Update virtual operands for
3443 the first and last stmt only. Add a dry-run capability.
3444 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3445 after CFG manipulations and in RPO order to catch all
3446 secondary opportunities reliably.
3448 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3451 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3453 2019-01-08 Richard Biener <rguenther@suse.de>
3456 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3457 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3458 virtual operand, also updating SSA use.
3459 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3460 Update stmt after resetting virtual operand.
3461 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3462 * gimple-iterator.c (gsi_remove): When not removing the stmt
3463 permanently do not delink immediate uses or mark the stmt modified.
3465 2020-01-08 Martin Liska <mliska@suse.cz>
3467 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3468 (ipa_call_context::estimate_size_and_time): Likewise.
3469 (inline_analyze_function): Likewise.
3471 2020-01-08 Martin Liska <mliska@suse.cz>
3473 * cgraph.c (cgraph_node::dump): Use systematically
3476 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3478 Add -nodevicespecs option for avr.
3481 * config/avr/avr.opt (-nodevicespecs): New driver option.
3482 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3483 "-specs=device-specs/..." if that option is not set.
3484 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3486 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3488 Implement 64-bit double functions for avr.
3491 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3492 --with-double-comparison.
3493 * doc/install.texi: Document them.
3494 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3495 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3496 <WITH_DOUBLE_COMPARISON>: New built-in defines.
3497 * doc/invoke.texi (AVR Built-in Macros): Document them.
3498 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3499 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3500 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3502 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
3505 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3506 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3507 when only building rm-profile multilibs.
3509 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
3512 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3513 lattice for a value to check.
3514 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3515 finite propagation in self-recursive scc.
3517 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
3519 * ipa-inline.c (caller_growth_limits): Restore the AND.
3521 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3523 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
3524 (VEC_ALLREG_ALT): New iterator.
3525 (VEC_ALLREG_INT_MODE): New iterator.
3526 (VCMP_MODE): New iterator.
3527 (VCMP_MODE_INT): New iterator.
3528 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
3529 (vec_cmp<u>v64qidi): New define_expand.
3530 (vec_cmp<mode>di_exec): Use VCMP_MODE.
3531 (vec_cmpu<mode>di_exec): New define_expand.
3532 (vec_cmp<u>v64qidi_exec): New define_expand.
3533 (vec_cmp<mode>di_dup): Use VCMP_MODE.
3534 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
3535 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
3536 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
3537 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
3538 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
3539 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
3540 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
3541 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
3542 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
3544 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
3545 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
3547 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3549 * config/gcn/constraints.md (DA): Update description and match.
3551 (Db): New constraint.
3552 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
3554 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
3555 Implement 'Db' mixed immediate type.
3556 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
3557 (addcv64si3_dup<exec_vcc>): Delete.
3558 (subcv64si3<exec_vcc>): Rework constraints.
3559 (addv64di3): Rework constraints.
3560 (addv64di3_exec): Rework constraints.
3561 (subv64di3): Rework constraints.
3562 (addv64di3_dup): Delete.
3563 (addv64di3_dup_exec): Delete.
3564 (addv64di3_zext): Rework constraints.
3565 (addv64di3_zext_exec): Rework constraints.
3566 (addv64di3_zext_dup): Rework constraints.
3567 (addv64di3_zext_dup_exec): Rework constraints.
3568 (addv64di3_zext_dup2): Rework constraints.
3569 (addv64di3_zext_dup2_exec): Rework constraints.
3570 (addv64di3_sext_dup2): Rework constraints.
3571 (addv64di3_sext_dup2_exec): Rework constraints.
3573 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
3575 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
3576 existing target checks.
3578 2020-01-07 Richard Biener <rguenther@suse.de>
3580 * doc/install.texi: Bump minimal supported MPC version.
3582 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3584 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
3585 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
3586 * langhooks.c: Include stor-layout.h.
3587 (lhd_simulate_enum_decl): New function.
3588 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
3589 handle_arm_sve_h for the LTO frontend.
3590 (register_vector_type): Cope with null returns from pushdecl.
3592 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3594 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
3595 (aarch64_sve::nvectors_if_data_type): Replace with...
3596 (aarch64_sve::builtin_type_p): ...this.
3597 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
3598 (find_vector_type): Delete.
3599 (add_sve_type_attribute): New function.
3600 (lookup_sve_type_attribute): Likewise.
3601 (register_builtin_types): Add an "SVE type" attribute to each type.
3602 (register_tuple_type): Likewise.
3603 (svbool_type_p, nvectors_if_data_type): Delete.
3604 (mangle_builtin_type): Use lookup_sve_type_attribute.
3605 (builtin_type_p): Likewise. Add an overload that returns the
3606 number of constituent vector and predicate registers.
3607 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
3608 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
3609 instead of aarch64_sve_argument_p.
3610 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
3611 (aarch64_pass_by_reference): Likewise.
3612 (aarch64_function_value_1): Likewise.
3613 (aarch64_return_in_memory): Likewise.
3614 (aarch64_layout_arg): Likewise.
3616 2020-01-07 Jakub Jelinek <jakub@redhat.com>
3618 PR tree-optimization/93156
3619 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
3620 least significant bit is always clear.
3622 PR tree-optimization/93118
3623 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
3624 simplifier with two intermediate conversions.
3626 2020-01-07 Martin Liska <mliska@suse.cz>
3628 * params.opt: Add Optimization for various parameters.
3630 2020-01-07 Martin Liska <mliska@suse.cz>
3633 * doc/extend.texi: Explain cloning for target_clone
3636 2020-01-07 Martin Liska <mliska@suse.cz>
3638 PR tree-optimization/92860
3639 * common.opt: Make in Optimization option
3640 as it is affected by -O0, which is an Optimization
3642 * tree-inline.c (tree_inlinable_function_p):
3643 Use opt_for_fn for warn_inline.
3644 (expand_call_inline): Likewise.
3646 2020-01-07 Martin Liska <mliska@suse.cz>
3648 PR tree-optimization/92860
3649 * common.opt: Make flag_ree as optimization
3652 2020-01-07 Martin Liska <mliska@suse.cz>
3654 PR optimization/92860
3655 * params.opt: Mark param_min_crossjump_insns with Optimization
3658 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
3660 * ipa-inline-analysis.c (estimate_growth): Fix typo.
3661 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
3663 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
3665 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
3666 helper function to return the valid addressing formats for a given
3667 hard register and mode.
3668 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
3670 * config/rs6000/constraints.md (Q constraint): Update
3672 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
3675 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
3676 Use 'Q' for doing vector extract from memory.
3677 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
3679 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
3680 doing vector extract from memory.
3681 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
3682 extract from memory.
3684 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
3685 for the offset being 34-bits when -mcpu=future is used.
3687 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
3689 * config/pa/pa.md: Revert change to use ordered_comparison_operator
3690 instead of cmpib_comparison_operator in cmpib patterns.
3691 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
3692 of cmpib_comparison_operator. Revise comment.
3694 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3696 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
3697 in an IFN_DIV_POW2 node to be equal.
3699 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3701 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
3702 (vect_check_scalar_mask): ...this.
3703 (vectorizable_store, vectorizable_load): Update call accordingly.
3704 (vectorizable_call): Use vect_check_scalar_mask to check the mask
3705 argument in calls to conditional internal functions.
3707 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3709 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
3710 '0' matching inputs.
3711 (subv64di3_exec): Likewise.
3713 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
3715 * config/mips/mips.c (vr4130_align_insns): Fix typo.
3716 * doc/md.texi (movstr): Likewise.
3718 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3720 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
3723 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3725 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
3727 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
3728 to a temporary file and use move-if-change to update the real
3729 file where necessary.
3731 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3733 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
3734 rather than Upa for CPY /M.
3736 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3738 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
3741 2020-01-06 Martin Liska <mliska@suse.cz>
3743 PR tree-optimization/92860
3744 * params.opt: Mark param_max_combine_insns with Optimization
3747 2020-01-05 Jakub Jelinek <jakub@redhat.com>
3750 * config/i386/i386.md (SWIDWI): New mode iterator.
3751 (DWI, dwi): Add TImode variants.
3752 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
3753 <general_hilo_operand> instead of <general_operand>. Use
3754 CONST_SCALAR_INT_P instead of CONST_INT_P.
3755 (*addv<mode>4_1): Rename to ...
3756 (addv<mode>4_1): ... this.
3757 (QWI): New mode attribute.
3758 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3759 define_insn_and_split patterns.
3760 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3762 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
3763 <general_hilo_operand> instead of <general_operand>.
3764 (*addcarry<mode>_1): New define_insn.
3765 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
3767 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
3769 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
3770 Use "call" instead of "set".
3772 2020-01-03 Martin Jambor <mjambor@suse.cz>
3775 * ipa-cp.c (print_all_lattices): Skip functions without info.
3777 2020-01-03 Jakub Jelinek <jakub@redhat.com>
3780 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
3781 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
3782 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
3783 for 'e' simd clones.
3786 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
3788 (mprefer-vector-width=): Add Save.
3789 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
3790 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
3791 (ix86_debug_options, ix86_function_specific_print): Adjust
3792 ix86_target_string callers.
3793 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
3794 (ix86_valid_target_attribute_tree): Likewise.
3795 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
3796 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
3797 ix86_target_string caller.
3800 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
3801 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
3802 instead of gen_int_shift_amount + convert_modes.
3804 PR rtl-optimization/93088
3805 * loop-iv.c (find_single_def_src): Punt after looking through
3806 128 reg copies for regs with single definitions. Move definitions
3809 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
3811 * config/arm/arm-c.c (arm_cpu_builtins): Define
3812 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
3813 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
3814 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
3815 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
3816 * config/arm/arm-tables.opt: Regenerated.
3817 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
3818 arm_arch_i8mm and arm_arch_bf16 when enabled.
3819 * config/arm/arm.h (TARGET_I8MM): New macro.
3820 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
3821 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
3822 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
3823 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
3824 (v8_6_a_simd_variants): New.
3825 (v8_*_a_simd_variants): Add i8mm and bf16.
3826 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
3828 2020-01-02 Jakub Jelinek <jakub@redhat.com>
3831 * predict.c (compute_function_frequency): Don't call
3832 warn_function_cold on functions that already have cold attribute.
3834 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
3837 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
3838 COMDAT group function labels in .data.rel.ro.local section.
3839 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
3842 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
3843 comparison_operator in B and S integer comparisons. Likewise, use
3844 ordered_comparison_operator instead of cmpib_comparison_operator in
3846 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
3848 2020-01-01 Jakub Jelinek <jakub@redhat.com>
3850 Update copyright years.
3852 * gcc.c (process_command): Update copyright notice dates.
3853 * gcov-dump.c (print_version): Ditto.
3854 * gcov.c (print_version): Ditto.
3855 * gcov-tool.c (print_version): Ditto.
3856 * gengtype.c (create_file): Ditto.
3857 * doc/cpp.texi: Bump @copying's copyright year.
3858 * doc/cppinternals.texi: Ditto.
3859 * doc/gcc.texi: Ditto.
3860 * doc/gccint.texi: Ditto.
3861 * doc/gcov.texi: Ditto.
3862 * doc/install.texi: Ditto.
3863 * doc/invoke.texi: Ditto.
3865 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
3867 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
3870 2020-01-01 Jakub Jelinek <jakub@redhat.com>
3872 PR tree-optimization/93098
3873 * match.pd (popcount): For shift amounts, use integer_onep
3874 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
3875 tests. Make sure that precision is power of two larger than or equal
3876 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
3877 instead of ULL suffixed constants. Formatting fixes.
3879 Copyright (C) 2020 Free Software Foundation, Inc.
3881 Copying and distribution of this file, with or without modification,
3882 are permitted in any medium without royalty provided the copyright
3883 notice and this notice are preserved.